If the statistical distribution of NPC stats does not conform to the rules presented for PC generation, then it creates a lot of problems. Invariably, it means that PCs don't play by the same rules as anyone else; and when the rules of the game reflect the reality of the game world, that basically amounts to codified meta-gaming (i.e. you succeed at the task because you're a PC; which rather cheapens any victory you might earn, when you know it's only because the rules are biased in your favor).
One of the few things I actually liked about 4e was that it got rid of needing to think about NPCs and PCs as needing to use the same rules, and harkened a bit back to 1e and 2e about only doing as much as was needed for NPCs. The opposite of that was my least favorite thing about 3.5/PF - why does a great Blacksmith necessarily get better at combat in terms of BAB and hp (as they do if you use the expert class)!?!
As far as ability scores in particular, the 1e DMGs words on it show there's a long history of the PCs and NPCs being different:
1e DMG said:NON-PLAYER CHARACTERS
Non-Player Characters: You should, of course, set the ability scores of those NPCs you will use as parts of the milieu, particularly those of high level and power. Scores for high level NPCs must be high -- how else could these figures have risen so high? Determine the ability scores of other non-player characters as follows:
General Characters: roll 3d6 for each ability as usual, but use average scoring by considering any 1 as a 3 and any 6 as a 4.
Special Characters, Including Henchmen: Roll 3d6 as for general characters, but allow the full range (3-18) except in the ability or abilities which are germane to his or her profession, i.e. strength for fighters, etc. For all such abilities either use one of the deterministic methods used for player characters or add +1 to each die of the 3 rolled which scores under 6.