Do you know the biggest differences in how the PER, Win Shares and VORP stats on basketball-reference.com? There's oddly wide variance in some cases and I'm not sure what that reflects.
PER is frankly not a very good stat. I hardly use basketball-reference anymore since NBA.com beefed up its stats. ESPN's +/- based stats have a better backing. But in general, stats per 100 possessions are more useful than stats--especially as possession stats become more accurate (based on actual touches that contribute to the result and not just an estimate of touches).
yeah there are a lot of great NBA stats that aren't really "advanced", just simple numbers but for people with an appetite for analysis that is more in-depth. but the most ultra-specific something is, generally the sample size becomes more problematic. and a lot of times, the more specific numbers just reinforce what we already know. i thought all those passing stats would be really awesome, and it turns out just looking at assists per game tells you a lot. its not perfect, and i still like the idea of looking at a points-per-assist attempt stat, or something similar (which i think you posted something similar recently, but i didn't check it out yet).
Sample size is definitely problematic in all stats. For example, it takes ~750 3Pt FGAs for the shooting percentage to stabilize. That's not saying they need that many attempts to reach the peak of their shooting, but that on average it takes that many attempts for 3Pt% to be a useful stat for evaluation. There's too much statistical variation based on luck when the sample isn't that big. To put that in perspective, that's more than any player has ever taken in a season. Source.