I've put an efficiency plot in attachment. This time it is really an 
efficiency - for all the MC muons with pt > 10 GeV, I look for a 
reconstructed muon (with pt > 5 GeV) within a cone of angle 0.1
  For MuonBox, there is still this 25% inefficiency between eta=1 and 
eta=1.3, already seen with 8.2.0. It cannot be related with DC1/DC2 
digit incompatibility, since the sample has been redigitized (and the 
efficiency from Moore is fine in this eta interval). 
  For Moore/MuId standalone efficiency is pretty good up to eta=2, then 
there is a drop of some 20%.   
  For the combined MuId, I get poor efficiency everywhere. I have used 
the same cuts of the dc1 note, fitcodecb=0 and 0<chi2mcb<10.

  Fake rates (defined as in the note):

MBox                 pt>10(25)     8.3% (2.9%)  was 4.2% (1.3%) in 7.0.3
MuId standalone      pt>10(25)    12.4% (5.6%)  same as in 7.0.3
MuId combined        pt>10(25)     2.4% (0.5%)  better than in 7.0.3

  Energy resolution:

                     7.0.3   8.3.0

MBox                  3.5%   3.6%
MuId Standalone       4.6%   3.9%      (good...)
MuId Combined         2.6%   2.6%