Two thoughts, to begin.
First, much of East Asia has extremely demanding placement tests for elite high schools (and sometimes lower schools). You might reasonably say that we're only taking students from good high schools around the world, but that's a comment about school quality, not students. As people often point out, Harvard dropouts are almost as successful as Harvard grads - merely getting in represents a powerful filter. So it's plausible that we're taking rich/educated elites worldwide, but in East Asia we've got a talent/skill filter comparable to college admissions being applied in advance.
Second, your point: there's a non-trivial penalty for being a Chinese applicant to a US college, and it's not particularly obvious what impact that has. I suppose it could be that the harsher requirements for acceptance are driving the difference, and we don't have any real evidence on how responsive outcomes are to changing that restriction. (Is there a good breakdown of outcomes for a given group by visa type? That might show whether nation-specific admission rules on some visa types are changing outcomes.)
(The third thought, of course, is "it's a huge multifactorial mess and there's probably not enough data to trust any conclusion we come to".)