I'd like to say that I'm still finding it an interesting problem. I was aware of the 26/65 =2/5 result from several years ago (it also works for 16/64, by the way) but your asking it prompted me to look into the problem seriously.
For example, it doesn't stop there, and indeed it seems that if AB/BC = A/C then you can add an arbitrarily large number of B's and you'll still get A/C at the end (in any base). I've not formally proved this yet (seems like a somewhat tough exercise), but heck, that's an interesting discovery.
In terms of the original problem, I think a general solution is possible: in base ten one finds that, if AB/BC = A/C,
10 = [(A-B)*C]/[(C-B)*A]
which implies that B must be smaller than both A and C, or larger than both. Incidentally, this also forces A, B and C to be different: if A=B then you get 0 on the right-hand side; if C=B you get infinity, and if A=C then you get 1, none of which equals 10. Hence A, B and C must be different.
I think there are ways to impose various other constraints on the values of A, B, C, eg I think you might find that A must be less than C.