In the 2009 analysis, we found that increasing the matching jet patch DSM threshold can improve the data adn simulation agreement. This is not well understood yet. So I looked at the reconstructed jet matched jet patch ADC in both data and embedding. In data, the jet patch has didFire()==1 and shouldFire()==1 requirement, and in embedding, the jet patch has only shouldFire()==1 requirement. All other conditions are the same in data and embedding. 
For example, in the following figure, left pannel shows with the JP1 or L2Jethigh requirement, the jet patch ADC versus jet patch index in data, and middle panel shows the jet patch ADC versus jet patch index in data, and the right panel shows the normalized jet patch ADC sumed over all jet patches in data (shown in red) and embedding (shown in blue). 
Figure 1, JP1 trigger Barel jet patch ADC.

Figure 2, L2JetHigh trigger Barel jet patch ADC.

As the embedding has only the shouldFire()==1 requirement, the below threshold distribution may related with the shouldFire()==1 but didFire()==0 case. Another interested feature is found by Brian, which the JP1 prescale may impact on the ratio of (shouldFire()==1&&didFire()==0) /(shouldFire()==1&&didFire()==1).
Further debug and check of all other jet patches is ongoing.