When we move from a nasal consonant /m, n, ŋ/ to a /f, θ, s/ or /ʃ/, we often accidentally make a /p, t/ or /k/ between them. This is because our tongue is making some kind of /p, t, k/ articulation left over from the nasal sound while the nasal cavity is simultaneously blocked for the articulation of the following sound. Such consonants are called epenthetic consonants. They are rarely included in transcriptions for words in learner's dictionaries. However, they often are included in transcriptions of words in more highbrow dictionaries and, importantly, in pronunciation dictionaries. In terms of guidance for learners it is probably best to aim for the version without the epenthetic consonant. If you happen to make one, fine. That's perfectly acceptable. It's quite easy to spot epenthetic consonants in such dictionaries as these optional/accidental consonants are usually in brackets within the word, or are written in superscript.
How we use our velums!
If you look into your mouth in the mirror and breathe out through your nose, you'll see a flap of skin/tissue hanging down at the back there and meeting the tongue, closing off the back of your mouth. If you then say "Aaaah" like you would for a doctor when they're trying to look at the back of your throat, you'll notice it lifting up off the back of your tongue so that you can see the back wall of your throat, a bit like a curtain that's being raised. It's got a little punch ball thing on the bottom of it, which is your uvula.
The name for that curtain-like tissue is the velum, literally veil in Latin. It plays a very important role in human speech. It's main job is to direct air either through the mouth or nasal cavity, or both. When you're resting it assumes a kind of half way position so that air can leave your vocal tract through both the mouth and nose during normal breathing. But if you make a consonant sound, for example the sound /t/, it raises up, blocking off access to the nasal cavity so that air is directed into the mouth. For a normal, canonical /t/, while this is happening, the tongue blocks the exit of the air by creating a seal with your alveolar ridge (that's the little shelf like part of your mouth behind your upper teeth). There is then a rapid build up of air pressure behind this blockage and when we release the /t/, we hear plosion which is the effect of this pressurised air being suddenly released. We would not be able to create this build up of air behind the /t/ blockage if we didn't have a velum. The reason is that all this air would simply escape through the nasal cavity and out of our noses.
In contrast when we produce a nasal consonant, such as /n/, /m/, or /ŋ/ the velum is lowered and the air passes out through the nasal cavity giving the sound a characteristic metally resonance. This time the air is blocked from leaving the mouth by your tongue and it is forced instead to exit via your nose. If you make a long /n/ sound and suddenly close your nostrils with your fingers, you'll find that the consonant is suddenly stopped. When you make a nasal sound the air needs to escape through your nose. Importantly, in terms of the Original Poster's query, when we make an /n/ sound, specifically, the tongue is blocking the mouth at exactly the same point it does for /t/.
When we speak our organs of speech are always anticipating sounds that are coming up a bit later. The smooth transition from one sound to another inside words, or across word boundaries, requires (subconscious) forward planning and small adjustments to the way we make individual sounds. The velum in particular, is slow to get moving, and it often starts it's journey upwards to close off the nasal cavity quite early in the proceedings, sometimes giving a nasal quality to vowels, for example, before a nasal consonant. So, if you listen carefully to the vowels in the words "hot" and "on", you might be able to hear that the vowel in the second word has a nasal tinge to it that is missing from the word "hot". This is because the velum is already anticipating the /n/ in this word and is already lowering, thereby letting air out of the nasal cavity giving the vowel that characteristic twang.
Now in the words conscious (in British RP, /kɒnʃəs/ in Gen American, /kɑ:nʃəs/), we see the sequence /nʃ/. What happens in this sequence is this: The blade of the tongue comes up to make a seal with the alveolar ridge for the /n/ blocking air from leaving the mouth. Meanwhile the velum is lowered allowing the free passage of air through the nasal cavity. The air coming up from the lungs, prevented from leaving the mouth, exits through the nose. Then, as the /n/ segment draws to a close and the /ʃ/segment begins, the velum raises to block of the nasal passage, the very tip of the tongue comes away from the alveolar ridge and the tongue furrows so that a kind of channel is formed down the central ridge of the tongue. The air now blocked from leaving through the nose is forced into the mouth and through this narrow channel. As it passes through this narrow gap it causes turbulence in the air heard as voiceless friction. In other words we hear a specific type of hissing noise.
Notice that there are two entirely different things going on here. On the one hand we have the tongue manipulating the air flow in the mouth, and on the other we have the velum blocking off or allowing access to the nasal cavity. In an ideal world the tongue and velum would work in perfect harmony so that the velum would block of the nasal cavity at the exact moment that the tip of the tongue moved away from the ridge. However, what often happens in reality is that the velum blocks off the nasal cavity a split second before the tongue comes away. During this split second, the air is directed into the mouth and there is a build up of air pressure behind the blockage that the tongue was making for the /n/. When the tongue does actually come away from the ridge this air is suddenly released. What has happened here is that we have accidentally reproduced the exact articulation for a /t/ sound: blockage at alveolar ridge, build up of air pressure, sudden release. This accidental articulation of an extra sound in such words is called epenthesis. We often see epenthetic consonants appearing when we have a nasal sound followed by a voiceless sibilant consonant (for example /s/ or /ʃ/). So we might see the following pronunciations for these words:
- mince /mɪnts/
- hamster /hæmpstə/
- angst /æŋkst/
These examples show the insertion of an extra /t/, /p/ and /k/ respectively.
So far I have described the insertion of such epenthetic consonants as accidental. Perhaps incidental might be a better description. These words sound perfectly natural with or without the extra consonants and many native speakers won't be able to hear the difference either way. Over the years many epenthetic consonants have become integrated into the formal spellings of words.
Should dictionary transcriptions of words such as these include versions with these epenthetic consonants? It is a good question. For learners it is probably not a good idea to give all the different possible versions of a potential word in any case. It is just not helpful. There is also the problem that if, for example, one deliberately accentuates the /t/ in mints, this would still be widely recognised as an acceptable form of the word. However, if we strongly accentuated a /t/ in mince this would be an unacceptable form of the word for many speakers (this is despite the fact that these two words are basically homophones for the majority of us). So for this reason a learner is probably best seeing the canonical versions of the words.
However, normal dictionaries and pronunciation dictionaries are an entirely different matter altogether. Here is part of what the world renowned Professor John Wells has to say about epenthetic consonants in his blog here. The abbreviation LPD stands for the also famous Longman Pronuncation Dictionary, which is probably the leader in its field:
Mariano was correct to make the generalization that plosive epenthesis happens (for those who use it) when ANY nasal is followed by ANY voiceless fricative, though only in certain syllabification environments. In my 2009 posting I mentioned fen(t)s, ˈkɒn(t)ʃən(t)s, ˈhʌm(p)fri, ˈgæŋ(k)stə (fence, conscience, Humphrey, gangster). So the nasal can be labial or velar as well as alveolar; the fricative can be any of f θ s ʃ.
In the examples just given the nasal+fricative sequence (or cluster) is always word-final or followed in the next syllable by a weak vowel: that is, these are cases where, in my syllabification analysis (which I know some people disagree with), the nasal+fricative cluster is syllable-final.
In LPD I do not allow for epenthesis in words where (in my analysis) there is a syllable boundary between the nasal and the fricative, e.g. ˌɪnˈsaɪd, kənˈsɪdə (inside, consider): I took the view that this absolutely blocks epenthesis.
The Original Poster's question
The long and short of this is that the Original Poster's suspicions are perhaps correct. An IPA phonemic transcription of the word conscious could be either /ˈkɑːnʃəs/ or /ˈkɑːntʃəs/ (RP: /ˈkɒnʃəs/ or /ˈkɒntʃəs/). This might be represented in a good dictionary as /ˈkɑːn(t)ʃəs/, where the bracket indicates that the epenthetic consonant may or may not be present.
A note on phonemic transcriptions
It has been mentioned in answers or comments on this thread that the transcription /ˈkɑːntʃəs/ should not be used with these slanty brackets. The assumption is that /ˈkɑːntʃəs/ is not a phonemic representation of the word. Presumably the thinking behind this is that there is no /t/ sound in the internal mental representation of the word. Although it is irrelevant, this is in any case probably a mistake, given that the most important representation here is one of sound not orthography. The fact that for many speakers some such epenthetic consonants are parts of the words concerned can be seen from the fact that such words are often misspelled using these very consonants (for example hampster). However, I think the premise here is misguided. A phonemic transcription of a word is a transcription which represents a pronunciation using specific symbols where each symbol represents a sound which if taken in isolation by a listener would be identifiable as a specific phoneme in that language. Now in the pronunciation of the word mince for example, there is often a sound which is clearly recognisable as an instantiation of /t/ by listeners. The fact that it is recognisable as this phoneme means that if we are giving a phonemic transcription of such a pronunciation of this word it is clearly a mistake to transcribe this as /mins/ and clearly correct to present it as /mints/. This is not a description of the quality of the sound and therefore no square brackets, , are required. The transcription consists entirely of a representation of (all) the discernible phonemes in such an instance of the word.