In 5e you prepare a certain selection of spells but it's completely unconnected to how many spells you can actually cast.
I don't have much to say in response to the other things (and I have already said far too much elsewhere), so I'll focus here.
How do you know what number of spells a spellcaster
should be able to cast? How do you know that the character
wouldn't be taught about these things by whomever taught them, or couldn't have figured out these things by doing (e.g. a Sorcerer experimenting with their abilities). How do you know that a spellcasting character definitely
doesn't have knowledge of how many spells they can cast each day?
If the game explicitly said that's how it worked, you might have a point, but it doesn't. That's the sticking point: there's no given reason a character
couldn't know that they can cast three level 3 spells once they're experienced enough. And there's no IRL comparison to make, so we can't do an analogue of "One Handed Catch" where we can bring intuitions in.
"Dissociation" when coupled with physical-world intuitions creates unfair limitations. We have no intuitions about magic because magic isn't real. You
have to have the game tell what the limits of magic are. Not so with non-magic. Non-magic is, implicitly, limited by what can be achieved in our world--even though we know people can and
do do things that can't actually be done in our world. That albatross hanging around the neck of characters that don't use "magic" proper
can be dealt with...but doing so in a way that never ever not even once invokes ANY form of "dissociated" mechanics is extremely difficult and (as you noted) may be impossible.
Magic gets a free pass on "dissociated" mechanics because it has nothing to "dissociate"
from. Which is, as I've said, part of why the whole concept of "dissociated" mechanics is flawed. "Dissociated" mechanics + intuitions about activities that resemble IRL ones
is the same thing as "it's not realistic," it just obfuscates the call for realism behind the unspoken intuitions.