Problems With The Albin

LOST badly today against Fide Master Terry Chapman, who came armed to meet my Albin Countergambit.
Chapman (2200) - Spanton (1927)
1.d4 d5 2.c4 e5 3.dxe5 d4 4.Nf3 Nc6 5.a3 Be6
This recommendation of Grandmaster Rustam Kasimdzhanov has been scoring better for Black than most major lines in the Albin.
6.Nbd2!
More popular is 6.e3, when 6...dxe3 7.Qxd8+ Rxd8 8.Bxe3 looks very good for White, at first glance, as he has got rid of Black's cramping d4 pawn without having to accept doubled pawns. But after, say, 8...Nge7 9.Nc3 Nf5, it turns out White has difficulties based on his uncastled king and the vulnerability of the e5 pawn. My record in this line, after 8...Nge7, is +5=2-1.
The text is stronger, I believe, and scores much better in ChessBase's 2018 Mega database.
6...Nge7 7.Nb3
White scores a very impressive 71% with this move in Mega 2018.
7...Nf5
7...Bxc4 was tried in Ivan Ivanisevic (2664) - Igor Khenkin (2655), 2008 Serbian Teams Championship. That game continued 8.Nbd4 Qd5 9.Qc2 Nxd4 10.Nxd4 Qxd4 11.e3 Qxe5 12.Bxc4, when White was a pawn up and had the bishop-pair in an open position (1-0, 44 moves).
8.Qd3 Be7?!
TC said in the postmortem he had been expecting 8...a5, with the idea of following up with ...a4 to displace White's queen's knight, and that does seem better than my choice.
9.Bf4 g5!? 10.g4 Nh6
Other knight moves are also unappetising.
11.Bg3 Nxg4?!
Probably better was 11...Bxg4.
12.Nbxd4 Nxd4 13.Nxd4 Nh6?
This retrograde move, coming after a series of indifferent moves, is definitely bad. I should have got on with development with 13...Qd7, although White is still much better.
14.Bg2 c6 15.0-0 Qd7 16.Ne6?!
Rather unnecessary as White's knight is at least as good as the bishop it captures. Bringing a rook to the d file made sense.
16...fxe6 17.Rfd1 Qxd3
My king is more exposed than White's, so I more-or-less had to swop queens.
18.exd3 0-0-0 19.d4 Nf5 20.d5 Nxg3 21.hxg3 exd5 22.cxd5 cxd5 23.Bxd5
After massed exchanges, we are down to a late-middlegame in which opposite-coloured bishops give some hope of drawing chances, but TC eventually wins in a surprising way.
23...Kb8 24.Kg2 Rhf8 25.Bf3 Bc5 26.Rxd8+
Also strong is 26.e6.
26...Rxd8
Can you guess White's next move?

Position after 26...Rf8xd8

27.Rd1!?
I was stunned by this voluntary exchange into an opposite-coloured bishops ending. It may well not be the best move in the position, but it does seem to be the simplest way to win.
27...Rxd1 28.Bxd1 Bd4 29.f4 Bxb2 30.a4 Kc7 31.Kf3 Kd7
Analysis engines Komodo9 and Stockfish9 give 31...h5 32.Ke4 h4 33.gxh4 gxh4, but reckon White is winning after 34.Bf3.
32.Ke4 gxf4 33.gxf4 a6 34.Be2 Ke7 35.f5
White's connected passed pawns cannot be stopped.
The game concluded:
35...Ba3 36.f6+ Kf8 37.a5 Bb2 38.Kf5 Bc3 39.e6 Bb4 40.Bf3 1-0