Hollinger puzzles me, because not only is he a stat geek, but he's not a particularly GOOD one. I mean, no pure statistical analysis is going to predict basketball performance. (Heck, by its very nature, statistics is DESCRIPTIVE, not PREDICTIVE ... that is, it models past performance, without making a claim as to future performance).
But, that said, there are a half dozen statistical measures that are significantly better/more accurate than PER. None are perfect, obviously, but Win Shares, Wins Produced, even simple Plus-Minus all have pretty good track records, which PER does not. PER is decent for explaining individual performance, but is heavily biased towards offensive production and scoring, and is just terrible at any sort of team analysis.
I just think if you are going to base your analysis SOLELY on statistics, why use the measure that every statistician in the business dislikes? (I mean, there have been lectures at Sloan about how bad it is...)