Functions, intertype relations, dichotomies, etc. Socionics does a lot to try to fit everything together in this interconnected labyrinth, and while that would be nice if everything happened match to up perfectly, I feel as if many of the nuances, deviations and factual insights are ignored in order in to maintain this nice neat interconnected matrix of ideas. Model A for example universally applies IEs to act in certain way depending on which function they in, but doesn't give us any insight in how exactly each type specifically uses each of the functions in their daily lives, or in what ways the same IE can manifest differently even when they are in the same function, which is something I have observed between Kindred and Look-a-like types.
The main point I am trying to make is that I think socionics should put more effort into going in-depth about each type, element, relation etc. and extracting insightful information rather then trying to maintain the model the information is in.