Xianglei created a new embedding sample where 10% of the primary track multiplicity is embedded into real data (originally, refMult was sampled, which does not make sense in FXT mode). This blog post serves as a comparison to show the difference by comparing two different pi plus samples.

First, a plot of the number of data tracks (tracks in the reconstruction which are not matched to an MC track) vs the number of reco tracks (tracks in the reconstruction which are matched to an MC track) from the original sample (produced on Jan 21, where refMult was sampled).There is very little variation in the number of embedded tracks.

Next, looking at the new sample using the primary track multiplicity
This has a much better distribution of the number of the number of reconstructed tracks, and will greatly improve statistics.

QA of other quantities appears to be good (aside from the persisting issue where the reco nHitsFit does not match the data distribution near eta=0, which is target rapidity)