Ha! Certainly not 
I’m definitely leaning toward models that are the best representation of reality possible. Now from a cost and pragmatic standpoint there is a reason why we are only using 3 primaries: it would be highly expensive to adopt display chains with more of them although it is the only way to optimally cover the visible spectrum and thus get a better model of reality.
Spectral rendering takes more time obviously, and if you aim to do it accurately, you ideally need to get your content creation pipeline setup for it which is more expenses.
I don’t think I follow you here! This is really a question of ratio between the components. If you want to see “more blue”, you need a higher ratio of blue light in respect to red or green, i.e. you need more higher frequency radiant energy.
Yes, the basis vectors are different and BT.2020 / ACEScg are producing better results, likely because the primaries are sharper along the fact that the basis vectors are rotated in way that reduces errors. A few people (I’m one of them) have written about that a few years ago about it. On that topic I’m re-hosting the tests Anders Langlands did in that respect: Render Color Spaces
Cheers,
Thomas