O'Kelly Variation - Game analysis

I lost this game eventually but did achieve a very favorable position out of the opening. I'm having a lot of fun playing the O'kelly though so I thought I'd share this and see if anyone has any thoughts. I would also be interested to see games from others in this sharp, double-edged, but always fun to play, variation.

Computer lines are the result of very precise calculations. But it's not necessary here. Black can simply play 10. Qd3 Nd7 (threatening Nc5 winning e4 therefore a better developing move than Be6) 11. Nfg3 Nc5 12. Qd2 h5 (to kick the Ng3) 13. h4 (forced but now it's a permanent weakness) g6 14. Nc1 Bh6 15. Qb4 Qc6. Black will finish development and/or bag e4 with an overwhelming positional advantage : two bishops, space, weak white pawns at h4/e4/c2.

But more attention should be put on the endgame. Black should not drift into a pawn down endgame and definitely not lose this.

6. Nf5(!?) is a white's inaccuracy. The best move: 6. Nde2.

I mean it’s well known that d4 is bad and that with c3 or c4 White can gain a nice advantage 

Thanks for taking the time to respond and thanks especially @Nerwal for the observations regarding Black's plans for the middlegame, which is what was lacking in the above. You mention space which draws my attention to 11...Bxf5 as an instructive mistake, trading pieces (B for N no less) while enjoying a space advantage. This led to future trades, and the freeing of White's position, since the diagonal (h1-a8) became open. The endgame required more care and patience. I have the Dvoretsky Manual as a chessable e-book and have built this into my study routine, along with some other endgame study resources). 

@Fizzy Seemingly not so well known at my level, although I'm sure I'll have to study lines you mention as time goes on. I've found that sometimes c3 transposes to French structures, which I'm comfortable with.