I don't think 3. Nb1 is that bad. It looks like an Englund Gambit declined (1. d4 e5 2. d5) with colors reversed. Black has an extra tempo, but the position is closed, so it shouldn't make much difference. It won't give you an opening advantage, but it should be playable.
END your 1.d4/London/closed game/positional woes forever with 1.d4 e5!? 2.dxe5 d6 3.exd6 Bxd6!
the compensation is that the position is easier to play with black, if you take the pawn I take back, kick the bishop out with f6 and castle and black is ahead in development
if not black will eventually take the pawn on e5
also +0.9 doesn't mean anything if you're not playing a GM or so (especially in the opening)
Here's how your position is performing on lichess for black at 2200 blitz. This is the position we reached without consulting an engine or theory, just playing good looking natural moves:

When you play suboptimal moves such as e3 for white your argument works out better, though still not very well... after Bxd6 white can just play Ne4 or Nc4 and try to trade, there's really nothing good you can do there beside trade, and the move is not that hard to find. At this point you have pretty much nothing. The logic for finding this move is kind of like.... well, I'm up a pawn, he's trying to attack so I'd like to trade down, he can't avoid the trade... Alternatively white can play c3 and plan to castle queenside. Castling into the attack via e3 would be low on my list of things to do here. As white I can see you planning to castle queenside / push your kingside pawns a mile away. Still, even e3 is performing well, though I think I'd probably have played c3.
Eval generally matters, it reflects the actual logic of the position. There are cases where human compensation makes up for loss of objective eval but it is not every case, this is not something to be assumed by default with no further analysis, to do that would be to allow Dunning-Krugers to rule decisions.
Generally speaking gambits work out better if a) you're white, b) the opponent plays some non-developing move, c) you're not throwing away central pawns, d) you have some complex pressure as a result. Here you have none of that. Even in the suboptimal position you setup - there is no serious threat on the d file. If white plays Rd1 your queen will be under pressure on the d file... Qd2 wasn't a super useful move. Your bishop is eyeing the kingside, but white hasn't castled kingside... and whites bishop is also over there. To push your pawns you'll have to retreat your bishop first. When you do tempo white bishop he'll just retreat it to g3 and threaten to trade it off for yours. White also has a pin via Bb5 he can play if he wants...
This isn't just abstract theorizing, I've played against the Englund many times and it is very noticeably easier to play against... it is really nothing like facing mainline d4 positions in terms of the difficulty.
I think the whole thing looks much more promising if you play the gambit with white.
Right, but after 2...d4 white's best line is to struggle achieving equality (3.Nce2 e5 4.Ng3 Be6!).