Changes leading to recovery of previously ignored muDst's

a) I forgot to remove the limit of 5K events per scheduler job (per run) , fixed, recovered 200K input events (order matters)

b) I had a typo in the hadd-macro and left out pattern P1. Previously you were seeing only AL for patterns P2+P3+P4. Now you can see all 4 fill patterns P1+...+P4.

c) I resolved the reason of sorter crash on this 13 runs,  due to bx7 vs.  bx48+offset being out of sync. 

  *for few low stats runs bx48 offset is wrong in spinDB , I inspected 2 runs, - runs discarded for now, spinDB needs fix, total ~250 input events.

  *for other very long runs this problem happened for the last event on muDst (to be clear not last W event, simply last event in the file) , I inspected 3 runs,  - I removed offending events and restore the runs.

The net gain is the change of spin sortable  input events from 0.99M to  1.46M.

 

Changes in monitoring of the spin dependent relative lumi to reduce possible bias from W-events themselves.

* for BHT3-rnd events used as primary relative lumi I require now that max DSM HT for given event is below 58.
* for QCD events used as controlle sample   I require now:
   vertex, high pT track, 2x2 EMC ET pointed by the track to be below 20 GeV. Those conditions reduce the # of QCD events by a factor of ~9.

 

 

Result:

The # of spin sorted Ws grew, errors are smaller, accordingly.

 AL(W+) stays at the same value, AL(W-) is now 0.7 sigma from zero, its sign did not fliped.