I *think* I read this in
- Signes-Codoñer, J. 2005. The definitions of the Greek middle voice between Apollonius Dyscolus and Constantinus Lascaris. Historiographia Linguistica 32: 1-33.
The Ancient Greek authorities (actually Roman-era) came up with something like 60 declensions for Greek, because they were not trying to do internal reconstruction or look for regularities. (I don’t know much about the Sanskrit grammarians, but what little I know tells me they were centuries ahead of the Greeks.)
The Latin grammarians did do internal reconstruction and looked for regularities. They got the Latin declensions down to five.
When the Greeks rediscovered Latin grammars in the Renaissance, they did a double take. Then, they took another, embarrassed look at their own grammar.
They worked out that with some pushing, they could get it down to ten.
With a lot more linguistics and reconstruction, we now have Greek declensions down to three; and if you’re aware of proto-Greek, the three make a lot of sense.
You can come up with more vowels, splitting off the contracted first and second declensions, and differentiating the third declension with vowel stems, which don’t look close to the consonant stems. If you do that, I’d be getting closer to 10 than 6: I’d want to break up several third declensions that don’t look obviously similar. (See Appendix:Ancient Greek third declension.)
If it makes you happier to think of βασιλεύς, -εως and τέλος, -ους as a completely different declension from πτέρυξ, -γος, because you don’t want to go via proto-Greek and Attic sound rules, well, you can *shrug*. People don’t do that, because Koine grammar teaching derives from Classical Greek grammar teaching: they use the same declensions, and just treat those odd forms as subclasses.