The "notch" in pseudorapidity is pronounced, and fairly precisely centered on eta=0.88 --

 chain->Draw("GlobalTracks.mEta","GlobalTracks.mNHitsFitTpc<=45&&GlobalTracks.mPt>0.4&&TMath::Abs(GlobalTracks.mEta/0.88-1)<0.1/0.88","") 

Plot number of hits fit vs eta (zooming in on 0.83 to 0.93 now...).  We are missing tracks with N fit points > 40 entirely in this region.

chain->Draw("GlobalTracks.mNHitsFit:GlobalTracks.mEta>>hh(100,0.83,0.93,46,-0.5,45.5)","GlobalTracks.mNHitsFitTpc<=45&&GlobalTracks.mPt>0.4","")

We zoom in again, to 0.86 < eta < 0.90, and this time plot the radius of the last hit on the track (a.k.a. the first point on the seed) vs eta.  Here we see that there is a class of events which is depleted for Rlast > 190 or so.

chain->Draw("rL:GlobalTracks.mEta>>hh(40,0.86,0.90,80,40.,200.)","GlobalTracks.mNHitsFitTpc<=45&&GlobalTracks.mPt>0.4","box")