no, no, the exception is that is A is attacked by C, but A is supporting an attack on C (by B say) then C doesn't cut the support. (that is a fairly simple case of A and B attacking C, it doesn't matter what C does, he can attack either of them, and it doesn't matter whic of A and B attacks and which gives the support, it is still 2 versus 1 so A+B beats C.
oh wait... nola mentioned that just more concisely...
Draug's exception to my exception is correct, but it means that if A is dislodged by an attack from C nothing A does can affect territory C, so if there is a seperate bounce(fight/conflict) in the territory C (say B attacking, and D or E attacking aswell) then the retreating A doesn't affect that fight, (even though normally the two on one situation would mean the territory C starts in was attacked by a strenght of two... this isn't a case of 2 vs 1, it's a case of B(1) vs whatever else is going into that territory, once it is determined that C has left winning his fight...
I think it is important that these exceptions makes some sense aswell. (and it is easier to understand them...