Starting from move 77 is correct until move 81.Kd3 Kf2 but 82.Bd5 ? it should have been 82.Ne2 Kg2 83.Be6 Kf3 84.Bf5 Kf2 85.Bg4 Ke1 86.Nc3 Kf1 87.Ne4 Ke1 88.Kc2 Kf1 89.Kd2 Kg1 90.Kd2 Kh1 91.Ke2 Kg1 92.Kf3 Kh1 93.Kg3 Kg1 94.Ng5 Kh1 95.Bg2 Kg1 96.Nf3 mate.
Bishop and Knight checkmate.

tablebase but of course there are different variation in there.
http://www.k4it.de/?topic=egtb&lang=en

I could see the logic in your method up to this point. The king was getting boxed into the corner, but then 89.Bh3+ allows Ke2 and the king escapes. (I used the same magnificent tablebase to check it.)
So 89.Kd2 is much better.
I'm interested in different techniques of doing this so appreciate your effort, thanks.
Last night at the club a friend told me of this method of a shrinking triangle formed by the bishop and knight. The king has to help too of course, because the knight only controls the 2 dark squares f2 and g3. But if the king covers e1 then together they can keep forcing the king into the corner. (The bishop on g4 prevents the king from getting to the last dark square, h4, by blocking the king's access via h3.)

Having done some more research, I see the method recommended by Capablanca is known as Philidor's 'W' method (getting the W from the knight moves.)
The other well known method is Deletang's Triangle which is what kco's tablebase sequence looks more like, (from move 87. onwards anyway.)
The thread from the forum, see here.
The actual game.....
http://en.wikipedia.org/wiki/Bishop_and_knight_checkmate