Miscellanea to be improved

[sarafiorendi]

• While working on the UL, I noticed that while in the PCL we discard the per-LS fits if sigmaZ is < 1 cm, we've never included this check into BeamSpotTools.
  By including it (not pushed to repo yet), we solve the problem of some "converging" fits with very small widths (X,Y,X) and large errors, which were seen, e.g., in
  https://indico.cern.ch/event/756915/contributions/3137469/attachments/1717113/2770610/slides_beamspot_SeptReReco_2018.pdf page 24.
  I have not understood why the fit is considered CONVERGED by Minuit though.

  • maybe it would be better to cut on the uncertainty value for the 3 widths, rather than on sigmaZ value

---

• Also, I think we should really change the track selection in the TkAlMinBias ALCARECO, because it's impossible to re-produce the original PV collection by refitting the available tracks.
  • on top of that, I noticed that when refitting the PVs we have these parameters of the PrimaryVertexProducer module that are set differently than from the "official" reconstruction (to be double checked):
    • TkFilterParameters.maxD0Significance = 5 [4 in official reco]
    • TkFilterParameters.maxNormalizedChi2 = 20 [10 in official reco]

---

Finally, here and here the range of the 1D fit has to be changed to -1,1.
I see the current settings cause a lots of failures in the fit (given that y is always negative).