[Disclaimer: this answer reports my own research on the subject]
This is the last vestiges of a change that took place during the Middle English period. In many languages, singular count nouns can be bare (without an article/determiner) when they are used as predicative nouns. After the copula, the types of nouns that can do this are usually restricted to nouns that denote professions. For example in French, you can say (1) but not (2).
1. Jean est avocat
'John is lawyer'
2. *Ce-ci est chaise
'This is chair'.
In Modern English, this use is even more restricted: you can only use a bare singular predicative noun if the role it denotes is unique. So we can say (3)
- John is president of that corporation
but if the corporation has a number of vice president roles, we can't say (4), and we certainly can't say (5).
- *John is vice president of that corporation.
- *John is lawyer.
Early Middle English, however allowed bare singular count nouns in all syntactic positions (subject/object and predicate). This is not surprising, since Old English didn't have an indefinite article. During the Middle English period (1100-1500) the use of these nouns dropped precipitously except in the "unique role" examples that are still possible today. By 1500, non-role predicate nouns were almost never bare. Non-unique roles were still possible into the early 18th century though, although they were quite rare. This is why we still find them in Early Modern English, so the Shakespeare examples are to be expected. But since by this time their use is quite rare, in Shakespeare we will find a mixture of both patterns. Since Don Quixote was published in 1605, I suspect that the translator has tried to give some of the flavour of Early Modern English to the translation.
[Source: Hanson, Schmitt and Munn (2014) using data from the Penn-Helsinki Parsed Corpus of Middle English (Kroch and Taylor (2000)]
The reasons for the loss are, I think, twofold: English lost grammatical gender in nouns, and at the same time gained an indefinite article (Crisma 2015, Curzan 2005). There seems to be a correlation between having grammatical gender and allowing bare singular predicates cross-linguistically (Munn and Schmitt 2005). The use of the indefinite article then became obligatory in non-predicate positions, and once gender was lost, its use became obligatory in predicate positions also.
Apparent bare singular nouns in comparatives
Even in modern English, bare count singulars like your first example He is more lunatic than he is knight are marginally possible. To my ear, they are much better with the phrasal comparative as in (6) than the clausal comparative (which has an actual verb) in (7). However the judgement seems to switch in the equative comparative, where the phrasal equative (8) seems worse than the clausal one (9).
- This hybrid animal is more tiger than lion.
- ??This hybrid animal is more tiger than it is lion.
- ??This animal is as much tiger as lion.
- This animal is as much tiger as it is lion.
I think that these cases are using the noun as a kind of mass noun which denotes the property of being the thing denoted by the noun. So (6) means something like this is tiger-like than lion-like. If this is the correct analysis, then these cases are not real count nouns and therefore like mas nouns and plurals, do not require a determiner.
Probably more needs to be said about these, however, since we can only use the property denoting mass interpretation in comparatives, I think.
Crisma, Paola. (2015). The “indefinite article” from cardinal to operator to expletive. In C. Gianollo, A. Jäger, & D. Penka (Eds.), Language change at the syntax-semantics interface. Mouton de Gruyter.
Curzan, Ann. (2003). Gender shifts in the history of English. Cambridge: Cambridge University Press.
Hanson, Kenneth, Cristina Schmitt and Alan Munn (2014) The loss of bare singular arguments and predicates in the history of English. Poster presented at the 16th Diachronic Generative Syntax Conference, Budapest, July 2014.
Kroch, Anthony and Ann Taylor (2000)
Penn-Helsinki Parsed Corpus of Middle English, second edition.
Munn, Alan and Cristina Schmitt. (2005). Number and indefinites. Lingua, 115, 821–855.