we avoid negativity during brainstorming to create psychological safety. that in turn expands the boundaries of creativity. the process doesn't end there however, which is what your criticism sems to assume, since we need to winnow down the ideas generated to a feasible subset, and then further to a pursued option. this latter part is where criticism (but not negativity) is employed to iterate toward the chosen option. literally no one argues the strawman that there can be no criticism during decision-making.
optimizing both divergence and convergence processes leads to good decision-making, not just one or the other.
To your point about psychological safety, that is actually reason to have an initial phase of idea generation occur individually before meeting in a group. Even a fully "positive-only" group likely inhibits divergent thinking to some degree.
Regarding groups, research challenges your claim that avoiding criticism during brainstorming "expands the boundaries of creativity", and that criticism is only useful for "winnowing down [already proposed] options." There are multiple studies showing that criticism during discussion _produces_ (not just selects) better ideas.
no one's arguing that we can't (or shouldn't) have any criticism during brainstorming, only that negativity (which can also be read as aggression) reduces psychological safety, which in turn reduces creativity. people do pull back (consciously and subconsciously) when the sting of (negative) criticism looms.
idea-focused criticism (rather than people-focused) can hone decision-making, but we still need that psychological safety to get the best outcomes. these things can both be true at the same time, and we can both succeed and fail at fostering either or both and not get the best outcome.
just as we might criticize an overly safety-oriented decision-making process, criticism-oriented decision-making processes tend to degenerate into contests of aggression, which tend to produce inferior outcomes. dogmatism in any direction doesn't help.
To put it in a nutshell: Criticism should be put forth in a "let's get this to work, together. Failure should not be feared; it's another data point" mindset.
optimizing both divergence and convergence processes leads to good decision-making, not just one or the other.