For 1. You need the article. Without an article, quantification in English is 'For all' or 'In general'. You are not comparing all red, green, and blue groups, everywhere, across all time. You are only comparing the ones relevant to this study. So you need an article.
For 2. You need the article. Again, no article means 'In general'. If you left it off, you would again be taking about all 5% significance levels, in general. That would at least be plural. Besides, you don't intend to discuss all, or even multiple, 5% significance levels, just the one you computed.
If you take the word 'level' out, you are breaking the idiom, and you need a different preposition. One can be 'at a level', but not 'at a significance'. The metaphor of a point on the measuring-stick as a place is lost.
You can say "With 5% significance or better" (not higher, lower, but that means better). But almost nobody does, because the field of statistics comes packaged with a set of formulaic idioms that all scientists usually stick with.
The more specific reason nobody varies from the narrow idiom, is that this choice of vocabulary does not stand up well to casual manipulation. Mostly because it really does not make a lick of sense. But it is traditional.
Everyone wants more significant results, but the number that represents significance is defined to be an estimate of the chance your data has misled you. So higher significance, in the natural language sense, means this number must be lower.
Also, you do not know the value of that number. You never computed it, you did computations to find out what values other numbers must have in order to ensure that that number was at most 5%. It may not actually have a value near 5%. It could be much lower.
But if it actually were lower, finding that out and publishing it would be unethical. Because it would imply you had either done this the hard way, with full intent to publish the exact confidence range, or decided ahead of time not to publish unless you got stellar results. And that would misrepresent your approach, and make your work hard to reproduce.
So it is just unwise to mess with the rigid formulaic way of talking about this particular statistic. The best thing to do is to not take the first step toward colloquialism, and just use the phrasing that the community has frozen in place, when you are writing.