Software is no different. So when someone claims that "you can't question the quality of AI output because quality of software is [partly] subjective", they're appealing to subjectivism. This position is generally held by the incompetent to entirely dismiss any question of craftsmanship or quality.
I've also never claimed that "you can't question the quality of AI output because quality of software is [partly] subjective". I'm not sure how you got there. I am simply saying that "quality" is an ill defined term and this is why we might see so much divergence when discussing AI code "quality".
You can absolutely evaluate code quality objectively, e.g performance, bugs, crashes.
What if it was rejected on an objective basis? You can't know, so you have to err on the side of caution, but you didn't.
So it's not embarrassing at all. In fact, it is more embarrassing to double down without knowing more.
It's a fallacy because you can always say that the choice of an objective metric was subjective, so everything is subjective.