Problem with track fit due to !695 and related MRs

The following discussion from !695 (merged) should be addressed:

• @sstahl started a discussion: (+3 comments)

  Looking at the refs, there seems to be something very wrong with the fitted tracks. The tracking efficiencies for BestLong are in the 40% range, they should be around 90%. In addition the tracking efficiencies of the Downstream tracking are lower by 20% comparing to https://gitlab.cern.ch/lhcb/Moore/-/blob/master/Hlt/RecoConf/tests/refs/hlt2_reco_baseline_with_mcchecking.ref.