Reading your post, I realize actually I use the thing in a more generalized sense as well, just as you've described. I've already been making those adjustments and operate more like the poker-player example. I can guess at motivations but there is no concrete certain indicator of everything, it's more a matter of building a complicated mental model in such a way that it's strongest / doesn't fall apart (i.e., conclusion by inference), and usually only general conclusions can be drawn. And new data always trumps theory; the theory has to be strong or broad enough to accommodate new input.
It's also why i shy away from typing people in general or typing public figures. Without extensive knowledge of their behavior + talking to them directly, there's no way to get inside the Black Box of Motivations and really know what drives the observable behavior.
I don't even understand myself sometimes.