I have to say that I did not understand the original question. However, trying to understand the intriguing solution given by belisarius, I came up with the following:
expr = (-1 + 1/p)^B/(-1 + (-1 + 1/p)^(A + B));
Together@(PowerExpand@FunctionExpand@Numerator@expr/
PowerExpand@FunctionExpand@Denominator@expr)
Conclusion (set by the velizari):

As an alternative:
PowerExpand@FunctionExpand@Numerator@expr/PowerExpand@
FunctionExpand@Denominator@expr
gives

or
FunctionExpand@Numerator@expr/FunctionExpand@Denominator@expr

Thanks belisarius for another good lesson in Mma's power.