Capa's Bishop

I think I have found a mistake in exercise 'Capablanca's endgame masterpieces, Capa's bishops'.

http://www.chess.com/chessmentor/view_lesson?id=6760

After black played Bc8, chess mentor says that Be6 doesn't win the g pawn. I didn't understand why not, so I checked this with an engine. The engine gives both Be6 and Be2 (and Bd1) about the same evaluation.

So this is a mistake in chess mentor?

1.Be6 leads to 1...Ne5

Whereas 1.Be2 wins the g-pawn immediately.

True, but after 1.Be6 Ne5 2.Bxc8 Nc6 3.Bxd6 Kxc8 white has won a pawn (and the game) as well.

According to Deep Fritz 13 1.Be6 is even slightly better than 1.Be2.

I think what yu need to keep in mind is that these are human players playing, not engines.  Black has an isolated pawn, and white plays a move that threatens to win that pawn. 

White has the Bishop pair, wins the g-pawn, has a passed pawn, and maintains dual threats. 

I agree with you. From a practical point of view it might be easier to win with 1.Be2.

Nevertheless, 1.Be6 is equally good. So it should be noted as such.