And you're of course correct that "fishes" exists as a plural form too.
Even if it were, just because "the meaning of something is obvious," it doesn't follow that the words are standard or even well-formed.
Native speakers of any language, and speakers of English likely even more so, generally have the ability to parse through malformed utterances of all sorts and recover most, if not all of the meaning, aoccdrnig to a rscheearch at Cmabrigde Uinervtisy. [1]
Besides, the meaning of many things is obvious even without the ability to comprehend the language. If I kick you in the ass and you start shouting at me in a foreign language I don't understand, I posit that the meaning of it would still be fairly obvious. We can even verify this experimentally if you are so inclined. Count me in for the test, I'd be glad to contribute my part for the advancement of science.
Regardless, any of the above does not invalidate my point that you cannot succesfully construct the standard plural "boxes" from "box" by following the pattern of "oxen" from "ox", and no amount of pedantry and nitpicking can change that.
However, even if you exclude two of the three examples I provided, then the remaining one still stands (and many more can be provided obviously). So I'm not really sure what is it exactly that you're trying to argue here. If it's just that my simple examples did not live up to your expectations, then I concede, and let's move on.
1. https://www.mrc-cbu.cam.ac.uk/people/matt.davis/cmabridge/
That's literally the linguistics definition of "standard or well-formed", so no.
> So I'm not really sure what is it exactly that you're trying to argue here.
Linguistics is a science, there are mathematical models that natural languages adhere to. Regardless of the social trappings around "standard"/acceptable speech, language really does follow mathematical laws. That's why these puzzles are meaningful and not just diversions for fun.
man -> men, pan -> ?