I’d beg to differ here. I think this mixes two things : unrealistic adaptation (mechanics-driven) and inaccurate adaptations (simplicity-driven). Unrealistic adaptation are deviation from what is possible to make the game richer and less fastidious; inaccurate adaptations are simplifications of the reality for either computational purposes or simply to avoid a ton of micromanagement, because this is unfriendly to most players.
The latter is something entirely acceptable by scientific standards. No model depicts the whole picture, and this is anyway not the purpose of a model. Its purpose is to depict and predict phenomenons of interest. That is nicely achieved by auto-evo (save for a few bugs), and all we have to do is increase the accuracy. Yet this does not equate to necessarily resorting to calculus: we are currently using time skips of hundreds of millions of years, and that allows for a lot of simplifications due to permanent regime. You may see an example in the end of part 1 for my compound diffusion model (which still has part 2 as a draft, by the way).
We can definitely work on the former, mechanics-driven adaptation, such as signalling agent which do not affect the player because we do not yet agree on how to make it enjoyable.
Why does this distinction matter? Because the reasons behind it are very different, and they should be treated accordingly. Simplified aspects requires theoretical and algorithmic input while unrealistic aspects need to find a proper gameplay mechanics brought forward by a game design perspective.