To check whether the prediction excess in the dimuon results below is due to mismodelling of the response of less pure muon categories, I reran the analysis using only CMUP soft muons. The results are below:
TCE mu+ mu+:
#mu^{+}_{l}#mu^{+}_{l}=0.0361493
#mu^{+}_{l}#mu^{+}_{h}=0.397905
#mu^{+}_{h}#mu^{+}_{h}=3.08399
nbg=3.51804
data=1
TCE mu+ mu-:
#mu^{+}_{l}#mu^{-}_{l}=1.26696
#mu^{+}_{l}#mu^{-}_{h}=0.268673
#mu^{+}_{h}#mu^{-}_{l}=0.537769
#mu^{+}_{h}#mu^{-}_{h}=13.4515
nbg=15.5249
data=4
TCE mu- mu-:
#mu^{-}_{l}#mu^{-}_{l}=0
#mu^{-}_{l}#mu^{-}_{h}=0.536577
#mu^{-}_{h}#mu^{-}_{h}=2.40745
nbg=2.94403
data=2
CMUP mu+ mu+:
#mu^{+}_{l}#mu^{+}_{l}=0.173812
#mu^{+}_{l}#mu^{+}_{h}=0.341771
#mu^{+}_{h}#mu^{+}_{h}=1.98673
nbg=2.50231
data=0
CMUP mu+ mu-:
#mu^{+}_{l}#mu^{-}_{l}=0.862434
#mu^{+}_{l}#mu^{-}_{h}=0.316661
#mu^{+}_{h}#mu^{-}_{l}=0.258145
#mu^{+}_{h}#mu^{-}_{h}=8.18717
nbg=9.62441
data=2
CMUP mu- mu-:
mu^{-}_{l}#mu^{-}_{l}=0.0179424
#mu^{-}_{l}#mu^{-}_{h}=0.3091
#mu^{-}_{h}#mu^{-}_{h}=1.55873
nbg=1.88577
data=2
CMX mu+ mu+:
#mu^{+}_{l}#mu^{+}_{l}=0.00847829
#mu^{+}_{l}#mu^{+}_{h}=0.139921
#mu^{+}_{h}#mu^{+}_{h}=0.884786
nbg=1.03319
data=0
CMX mu+ mu-:
#mu^{+}_{l}#mu^{-}_{l}=0.406995
#mu^{+}_{l}#mu^{-}_{h}=0.270739
#mu^{+}_{h}#mu^{-}_{l}=0.198187
#mu^{+}_{h}#mu^{-}_{h}=5.45729
nbg=6.33321
data=1
CMX mu- mu-:
#mu^{-}_{l}#mu^{-}_{l}=0.0243713
#mu^{-}_{l}#mu^{-}_{h}=0.2176
#mu^{-}_{h}#mu^{-}_{h}=1.30038
nbg=1.54235
data=0
The background prediction is still high everywhere and is very high in the opposite sign. The problem does not appear to be related to the soft muon type.