Well, after 12.Nd5, White's most active piece is very strongly placed, and retains a firm hold of a share of the centre. After 12.Nxa8, every White piece ia on an edge square; Black has just about exclusive tenure of the centre of the board. Taking the rook is an "OK, then, show me!" move.

Let's evaluate the position objectively: it's an open board; White has zero development, and his king is stuck in the center. Black has a centralized knight on a good outpost, a huge lead in development, both bishops have good diagonals available to them, and he can castle artificially by playing ...Kg7 getting his other rook into the game (with tempo, if the Bc7 has moved). As White, I would not play Nc7+/Nxa8. From a positional viewpoint alone, I would consider it too dangerous. Tactically, after the obvious 12...Bb5+/13...e4, I don't a safe spot for the white King.

12 Nd5 Qh4+ 13 g3 Qh6 14 Qc1 Qxc1 15 Rxc1 Be6 again is hardly a success for White

I would say the last mistake. Bd3 was advisable, in my opinion.

Hityerclock is right, but any player likes gaining material whenever possible.

You can see how wrong is your evaluation: his king has no threat, yours has 4 pieces "looking" at it (B, N, Q, pawn AND quickly after the lsb).

This position is given in Jonathon Rowson's excellent book on the Grunfeld. "13...e4! seems to provide excellent compensation for the material and I'm not sure how White shepherds his king to safety. This follows analysis by Lechtynsky, a chap I have never had round for afternoon tea, so it's worth checking it over as Black is somewhat short of a rook but nonetheless my impression is that White has problems here.".

Yep, this was a reference source I used in the game, I also believe Black is already doing well here

Incredible that such a position was even considered in a book... and incredible that serious study can give you to a book reporting it! I admire that...  

This position was in a book. Wow!

Rowson only gives e.g. 14 g3 Bc5 15 Kg2 Nf5 "when the initiative persists and my only objection is that the knight on a8 appears to be snoring." This position must be critical. Also possible were: 14 a3? e3+ Kxe3 Qe5+ 16.Kf2 Bc5 intending ...Nf5 wins 14 Ne2? Nxf3! wins 14 Qc1 Nf5 15 Ne2 exf3 16.gxf3 Bc5+ 17 Ke1 Qh4+ 18 Ng3 Kg7! 19 Nxg3 hxg3 20 Qxh1 is not all forced, but again looks very good for Black.

It is also worth adding that White doesn't have time to start to bring the Na8 back into play. 14. Nc7 e3+! 15. Kxe3 is forced then Nf5+ if 16. Kd3 Qd4+ 17. Kc2 Ne3+ 18. Kb3 Nxd1 and black is winning easily. White has alternatives, but none of them are any better, for instance 16. Ke3 Ng3! 17. hxg3 18. Qe5+ and if Kd3 Bf5 mate or Kf2 and Bc5 mates.

16. Ke3 is impossible (the king is already on e3). You don't mean Ke4 as then Qe5+ would give away the Q. I'm not sure what you do mean  

or 15 a3 Bd2 16 Qxd2 (otherwise Qd4+ wins) e3 17 Qxe3 Nxe3 18 Kxe3 Qxb2+ 19 Rd1 Kg7 20. Kf2 Bf5 -+ I don't see anyway back for White after 15 Qd5, considering the next moves are forcing, which seems to put 11.Nc7+ under a cloud.

There had to be a reason why the pawns are shaped like pacifiers!

I really don't know.

Maybe the engine assume white will play Rd2, but after Be3, white looks dead (say Qc2, BxR+, QxB, Qa1+, and if Qd1, Qxa2 threating c4, and if Bd1, Bxc4 at once)