9-Jan-2003  loss.breakdown.2

Breakdown of the missing BMU stubs:
There were 128 events with a Z leg which supposedly should have had
a BMU stub tacked on.  I tinkered with the stub-finding algorithm to
fix a bug, and added stublet-finding (2 hits near wire + BSU).  The
stublet finding is not tuned.  I then scanned the file using evd.

I corrected a bug in the BMU code.

	The new results:

39	BMUS is found now
12	Stublet is found now
___________________________________________
51	Stub is retrieved (Yay!)


 0	BMUS not found, should have been (usually 3-hit)
 4	Stublet not found, should have been
___________________________________________
 4	Problems in algorithm or tuning


28	Impossible to reconstruct.  Usual form is hits in 2 layers only
 2	Stublet would have been found with different parameters
 2	Not sure:  sometimes no BMU hits at all
___________________________________________
32	Not reconstructable:  delta-rays and chamber malfunctions


 5	CMX only, no IMU
25	CMX, skims the IMU (picks up BSU and perhaps outer layer BMU)
 8	Misses the short chambers
 3	A stack is not actually present
_____________________________________________
41	Fiducial issues:  IMU track not actually possible


	I don't have the original numbers, but from the claim that we
were overall 61% efficient I gather there were 328 events in the original
sample, of which 200 had a reconstructed BMUS.
	I subtract from 328 the 41 events which miss us, and add to
the 200 the 51 events newly retrieved.  This gives us 87% efficiency
overall, with the possibilty of achieving 89%.