I have done a little Maple study on efficiency concerning the 
a-parameter. The purpose of the exercise is to get a feeling for the 

I have tried to model the efficiency dependence according to the BAD. 
If I fold that distribution with two Mx distributions that roughly 
agree with what is caused by the variation of the a-parameter I can 
study the variation of the number of events below 1.55 in Mx.

The original model distributions differ by <7%. After folding with the 
efficiency the effect translates into a 20% variation in the number of 
events and hence on the branching ratio. Surprisingly the largest 
variation is at the rho-peak.

It is thus important that in all these studies the differential 
efficiency is properly applied. Incidentally do you have the efficiency 
distribution for the non-resonant model? I am worried that what is 
called efficiency may be a convolution of efficiency and resolution.