Last Round Melt Down

For as long as I have been playing chess, the last round to win position has always been disastrous - I mean simply not do-able. 2000 Harvard Open, I was 3/3 and lost, 2004 G/60 Champs I was 3/3 and lost, at the 2004 Northeast Chess Summer Getaway I pulled it out and won, but that was just the once. Since then, 2005 G/60 Champs I was 2.5/3 and drew, MCC Dec 2005 I was 3/3 and lost, 75th MA Open I was 4/5 and tied for first and I drew, and then 2006 RI Champs I was 4/4 and lost. Sometimes, like at the RI tournament, I was the major favorite and even had the white pieces, and still managed to lose in a game that I only had to draw!
I don't know what it is, call me a choker, but I never seem to be able to pull it out at the end. This week at MCC, I had 2.5/3 and needed to win for a share of first, and I had the black pieces. Not the most auspicious set of circumstances, though it's quite nice to come back after a break and go through three rounds without a loss, but arghh, I could feel it coming again. So I can't exactly say that I pulled out all the stops, but I thought long and hard about what I would play and what my strategy would be, and for a while there it seemed to be going alright.

Edward Astrachan (1866)-Joshua Haunstrup (1880) after 22...Bc4
Black has some active ideas from here, and white's center is likely to become more of a liability than an advantage. Certainly, it's not the kind of raging crazy position you would generally pick for a must win game, but I thought it had potential.
23.Rad1
and now I got to thinking. It seemed that the best thing to do was to keep him on his toes. I looked at 23...Ng3 with something like 24.Rfe1 Bxd3 25.Rxd3 Nxe4, but this seemed uncertain and dubious because of the revealed threats. I think I even considered 23...Nf6, but that certainly wasn't any good. Finally, I settled on
23...Bb3 and the game proceeded
24.Rde1 Nc4
25.Bc1 Rad8
26.e5
and now I was beginning to unravel. Sure, I had looked at this possibility plenty of times, but I always seemed to think that I could get my knight back to f6 to move to d5 supported by the other knight first. It was one of those things where I had considered it a long time ago in my thought process and then just assumed that it was a constant - as if the shifts in the position elsewhere weren't affecting the evaluation of this idea. I guess I could now have played something like 26...c5, continuing my activity, but I was growing afraid of g2-g4, which perhaps is a ghost, as it will take white a while to gobble the knight, and meanwhile, his position may collapse, but needless to say, I went in for:
26...Nxe5
27.Nxe5 Bxe5
28.Rxe5 Rxe5
29.dxe5 Rxd3
30.Bxh6
and when the dust settled, I found myself looking at a draw. Yeah, that's right, another no go fourth round.
I think that the best line was where I chickened out on move 23 and should have played:
23...Ng3
24.Rfe1 Bxd3
25.Rxd3 Nxe4
26.Bxh6 Nxc3
27.Bxg7 Rxe1+
28.Nxe1 Ncd5
29.Bh6 Re8
With what should be a steady advantage in the endgame. But I was lazy I suppose, and didn't calculate carefully enough... Someday I'll learn to pull one of these things out.