...
So that leaves us with 216 to consider.
Of these 134 truly missed any active material in the ACD.
*64 of the remaining events did have at least one MC "hit" within a tile, but the tile was determined to be below threshold by the digitization algorithm.
18 had MC "hits" in ribbons that were determined to be below threshold in the digitization algorithm (though we should note that the algorithm for the ribbons needs adjustment - namely in terms of the mean number of PEs by position along the ribbon)*
Of the 134 that failed to go through any active material in the ACD:
118 slipped through the gaps along the corners where the side tiles meet.
and
16 went through between two columns of side tiles on the same face, but didn't' seem to go through a ribbon.
Conclusions
The current geometry model in Gleam is more leak prone than the Geant3 model. The leakage through the side corner gaps is about 3-4 times larger than what is seen in the Geant3 model. Alex suggests that we look at how large those gaps are and check the distance between the ACD side tiles and the sensitive area of the TKR (current the G3 simulation is 8.5 cm).
Since the ACD in Gleam can be considered worst case, updating the ACD geometry will wait until after DC2.