The problem with ASR is that they’re findings are limited to the capacity of their instruments to perceive what will sound good to humans.
But the opposite approach of turning to experienced reviewers’ subjective opinions is no more reliable.
We’d benefit greatly, imho, from scientific methods applied to the listening side of the equation. A marriage of the two different, and both equally unreliable approaches we have today, in hopes of gaining more clarity.
Of course personal taste is a wild card, but at least we’d be a bit closer.