That's not an assumption, any more than it's an assumption that one can compare goalies with any metrics. I don't know what you mean by "we can't actually measure the expectations, so there is no way to confirm if any of the various formulas usually produce the correct answer." What would the "correct" answer be?
It seems that you're saying that because there is no definitive "good_goalie" response variable, GSAx is also not a good metric to measure goalie performance with. Because if there were a "good_goalie" variable, you could just see if GSAx is strongly correlated with it, or at least more strongly correlated than traditional metrics.
It might be true that GSAx is a poor proxy for "good_goalie", but saying that it makes fancy calculations and assumptions really just saying there is no way to measure goaltender performance, as metrics such as save percentage, GAA, Wins, etc. are undeniably more flawed, and also have no "good_goalie" to compare to.
I frankly don't see why it would be any worse than the eye test, which is famously poor and enormously subjective. If a goalie plays well according to a person, and it produces no tangible evidence, then that eye test is really just vibes.