I don't think it's a crappy stat in gauging players who average 30 minutes or more. Look at the top-20:
View attachment 45352
I don't see any outliers there except for maybe Harrell; but he's 2nd to Gobert in TS% this season so his efficiency bumps his PER up. There's no stat noise in the top-20
I don't even think there is much noise in the next 20 PER rankings:
View attachment 45353
but you will start to see players with lower minutes (Love, Robinson) who probably shouldn't be ranked. I thin PER normalized for minutes is a decent stat, but only so far. Getting down around 25 minutes introduces too much variance. PER actually does a good job of matching eyeball tests for the top 40-50 players. Lower than that, not so much
Its biggest weakness is with players who are good mainly for their defense. All box score stats have trouble measuring defense because they don’t have much to go on. Also players can get steaks, blocks, and rebounds while not being tirr the ally sound in defense.