I ran 1000 FAKE_EVENT 20 GeV Pt negative muons through GEANT. Of these, 74 failed to hit the chambers correctly. 84 times there was no stub found. I am investigating: this seems an unacceptably high failure rate.
There is a difference between the direction of positive and negative muons in the Bmu system, as tracked by GEANT. The first page of this postscript file shows the difference between its direction angle and its position angle, which reflects the deviation from radial direction. The second page shows the X/Y position of the missing tracks. These line up very nicely on the whole with the position of the gaps. I conclude that the inefficiency is largely geometric. I require 3 hits or more for a track, and a radial track which passes through one inter-cell gap is likely to pass through a second one, leaving only 2 hits behind.