Each player/GM/group's Miles May Vary, so it's probably best to agree to disagree on whether GMPC's are good or bad. Argument is degenerating quickly....
It's never best to agree to disagree (a false surrender).
@Shadowjack. You've just described (in further detail than before) the symptoms of what I'll call a sophomore GM. I'll spare the academics of competition theory and leave it as, you describe a game and a preference for games which shows a distinct lack in certain skillsets necessary to a high quality GM, among which are game balance (no GMPC should ever be
needed if your GM can competently write custom modules for your PCs) and table managements (2-3 players? A bare bones table is generally 4 players and a GM, a skillful GM will be able to run maybe twice that, I draw my line at 6 (8 for D&D but that's second nature to me now)).
What we're looking at here is the contradiction of mystery. For those who understand no explanation is needed, for those who do not no explanation will suffice.
I suggest you sit down with some game designers (I don't mean the writers of an RPG, I mean the guys that design games from nothing into a set of rules and actions) and ask them specifically to explain competitive models, the 4th wall, and metaplay to you.
btw, the best GM I've had the privilege of playing with has been Ed Greenwood. He used to run an irregular game at the Game Guild in Lake Geneva and I had the privilege of sitting in after being eliminated early in a M:tG tournament. After the immortal Ed, there's a whole pack of expert GMs that I have played under (including Chris Tulach and other RPGA and Judges Guild bigwiigs), many of whom aren't just top notch convention GMs and writers, but literally have given seminars on how to GM effectively (I'll never forget the KISMIF anbd Illusion of Free Will seminars from classic GenCon in MKE).