Benoni? No, Morra Gambit!

PLAYED in the Battersea Club Championship last night.
Spanton (168) - Paul Stokes (158)
Delayed Morra Gambit Declined
1.d4 c5 2.e4 cxd4 3.Nf3 Nc6 4.Bc4 Nf6 5.Qe2
Not 5.e5? Nxe5 6.Nxe5 Qa5+ etc.
5...d6 6.0-0 g6
6...Bg4 is probably more critical, when I have tried 7.c3 and 7.Rd1.
7.Rd1 Bg7 8.c3
Finally offering the c pawn, and so making the opening a Morra Gambit by transposition from the Benoni.
8...0-0
On 8...dxc3 I planned 9.e5, although 9.Nxc3, reaching a standard Morra set-up, is also playable. After 9.e5, Stockfish10 and  Komodo10 give 9...cxb2 10.Bxb2 Nh5 11.exd6 0-0 12.Bxg7 Nxg7 13.dxe7 Qxe7 14.Qxe7 Nxe7 15.Nc3, when the engines do not believe White has enough compensation for being a pawn down.
9.cxd4 d5
Stockfish10 prefers 9...Bg4. Komodo10 for quite some time rates both moves equally, but also comes to prefer 9...Bg4.
10.exd5 Nxd5

How should White proceed?

*****
*****
*****
*****
*****
11.Ne5!?
I rejected 11.Nc3 because I thought Black could grab a pawn, after 11...Nb6 12.Bb3, by 12...Nxd4?? 13.Nxd4 Bxd4. But I completely misjudged the position as White has, for example, 14.Bf4, when Black cannot save his dark-square bishop.
I also seriously considered 11.Bxd5!? Qxd5 12.Nc3, with something of an initiative, but I was reluctant give up the bishop-pair, especially when it meant exchanging my better bishop (the engines reckon the position is dynamically balanced).
11...e6
The engines prefer 11...Be6. The problem with the text is that Black will have difficulties developing his light-square bishop.
12.Nxc6 bxc6 13.Nc3 Qc7
Komodo10 is happy enough with this move, but the queen will not be safe on c7 once a white rook reaches c1.
14.Be3 a5
On 14...Nxe3 I planned 15.fxe3, but the engines point out 15...c5, when the position is opening nicely for Black's bishops. So the engines prefer 15.Qxe3, with equality, according to Komodo10, but Stockfish10's assessment of an edge for Black looks nearer the mark to me.
All of which suggests White's dark-square bishop should have been developed elsewhere, eg at g5.
15.Rac1
Threatening to capture twice on d5.
15...Qe7 16.Ne4 a4?!
...Nxe3 was still a decent move for Black.
17.Bg5 Qa7?
I expected 17...f6, when White must be better.
18.Bxd5 cxd5?
Black is holding on after 18...exd5 19.Nf6+ Bxf6 20.Bxf6 Qd7, although c6 is a chronic weakness and his kingside dark squares are a major concern.
19.Nf6+ Kh8
19...Bxf6 20.Bxf6 is clearly hopeless for Black as his queen is cut off from the kingside.
20.Rc3 Ba6
The engines give 20...e5!? 21.dxe5, but with a winning advantage for White.
21.Rh3 h6 22.Qf3 Qe7 23.Bxh6+ 1-0