For RT ticket 2106,  there was some concern that the database updates between 20100821 and 20101213 might be somehow be responsible for the difference in Nfit distributions seen between embedding and data. So I just processed 1000 events from st_physics_11044016_raw_4040001.daq using SL10k with the same chain as used in the Run 10 200 GeV AuAu production twice (once with DbV20101213 and once with DbV20100821). Here are the Nfit distributions of the two for global tracks with flag>0 (black is DbV20101213 and red is DbV20100821):

They look nearly identical to me. Here are the same distributions vs. track eta (DbV20100821 first, then DbV20101213):

...and vs. phi of the first hit on the track (where [-360,0] is for z of the first hit < 0, and [0,360] is for z > 0):

Again, the distributions look quite similar between the two DbV timestamps, with the exception of the disabled sector 20 (phi of [-45,-15]) with the newer DbV.

Unless I need to apply some other cuts, I see little reason to suspect the database updates are the issue.

-Gene