So the bias in the ATP stats, when they're wrong, is toward big servers like Sampras, Becker, Ivanisevic.
Not just their total points won, but their service percentages and success on first and second serve, are going to appear higher than they actually were.
I think I've read on this board a few debates about whether Agassi or Hewitt had the better return game. That's a good question, but if you use their stats against Sampras, it's more likely that Agassi's return stats are going to appear lower than they really were. Hewitt's stats are probably more accurate because I don't think I've found any ATP stats that are incorrect since around the year 2000. I do sometimes see the ATP differ with other sources by tiny amounts, but that's all. It's in the '90s that, so far, I've found all the big mistakes.