Yup, exactly, the knight is not best placed on f6 (in the exchange variation), so it's worth losing a tempo for this. Of course, the extra tempo also probably makes the knight placement more playable.
But ...dxc6 is not dubious! It gives black a faster development and his game is much less awkward, although ...bxc6 isn't horrible or anything, it's quite inferior to the normal capture.
Inferior how? The way I see it, Black needs the d-pawn to support the e-pawn and the center, otherwise the DSB has to assume the responsibility of doing so.
...bxc6 allows white to play a quick d4, so I don't think black will be able to have both pawns in the center. He'll probably have to give up the e pawn for the white d pawn, meanwhile, he can play for ...d5, but his pawn structure is weaker than usual in the exchange and he's not really ahead in the center, plus his development is slower. Compare that to when the bishop is on d6: black actually wants white to open things up but if he doesn't black will be fine in the center. White may even be able to play exd5 and after ...cxd5 try to prove the center pawns weak, or just leave the pawn there, maybe push it to e5, but white is going to have some potential pressure on the pawns. As white I would probably play 6 0-0 first (threatening Nxe5) and then d4, but with the idea of recapturing with the queen since it should be safe there. It would probably be a slightly favorable version of the scotch game. I just don't think in actual practice black can get his dreams of occupying the center before white does. I should probably post some analysis of what might happen.
I suppose it's better though than it is in the real exchange variation, as the attack on e4 could prove useful. I used to think a long time ago ...bxc6 would give better central control but really it just slows development and although he might get in ...d5, not with the e5 pawn still there and his structure is attackable. While it's true black's dark squared bishop is passive, it deters white from playing d4. If white plays with d3 black will not be outmanned in the center and his development is smoother with the extra tempo. The only problem with black's game of course is white's kingside majority. Black holds his own in the closed centers with d3, with the idea of playing ...f5 eventually to open some things up. It's probably equal, I mean, how strong do you expect bishops to be in a closed position anyways? But black can, eventually open things up possibly.
Well, elubus, I do agree with you to a point, but I am finding that of the ten or so games with ...bxc6, top-level chess players are among the ranks. Specifically, Keres and Botvinnik both scored wins with the line. Of course, to be thorough, I must also mention that Karpov, the "king of position" preferred ...dxc6 in his encounter with the Bayreuth Variation. I will try to get links to the 2 games I found where Black decides on ....bxc6, not because I want to prove that it is better, but because the resulting positions are very unique to the specific variation of the Ruy Lopez and help demonstrate the diversity of the particular opening. I think you'd find them interesting.
How many games were played with ...dxc6?
Anyways I actually used to play ...bxc6 when I sucked, which seems to support ...d5, but things are not so simple and the pawn structure has potential to be weak unlike with ...dxc6, plus development is hard as in many other e4 e5 openings, but the exchange is supposed to be the one where black's pieces are free.
When I first saw this variation, I was a bit surprised, and for obvious reasons, too. The Bayreuth Variation goes as followed:
It baffled me that White would retreat the bishop on move 4 only to exchange on move 5. Time-wise, this appeared very wasteful. White moves the Bishop 3 times in order to capture the knight on c6 that had only moved once. The capture could have been done on the previous move, but instead White waits for Black to develop a knight on f6 first.
Why does White do this? The answer appears in many games with this opening: Position. In the normal Ruy Lopez Exchange, the Knight on g8 is mostly deployed to e7 and from there to b6. The f6 square is then left open for pf7-f6 to support e5. The Bayreuth Variation leaves Black's position a bit less desirable; not only does Black have the doubled pawns on the c-file that occur in the Exchange Variation, but he also has to deal with his center pawn on e5 being a backwards pawn.
Of course, many will not be convinced by simple analysis to use this line. In such cases, master games serve as great examples of these ideas.