Frank explains that the theorists have this preference for a linear addition for various reasons but since the true sources are different (and e.g. the m_top dependence is very variable with pTH) he thinks they should be uncorrelated parameters in the final prescription.
Subsequent discussion on the synchronisation:
There are some slight differences in the detail of how CMS and ATLAS derive their high-pTH migration nuisances. The ATLAS numbers seem to be slightly larger - CMS is happy to change to these numbers, although it could be useful to discuss in slightly more detail.
The overall magnitudes of the uncertainties applied to each bin seem consistent between ATLAS and CMS, although the exact nuisance parameterisation scheme differs. CMS has kept e.g. the yield and res nuisances, and has a migration nuisance at 200 which is slightly different from ATLAS’ in that it (very slightly) affects the bins below 200 also. CMS has a generic boosted nuisance motivated by the old m_top scheme - this can easily be replaced with the updated calculations for the m_top nuisance as presented by ATLAS. Finally, ATLAS proposes a single nuisance to account for the pT(Hj)/pTH boundary, whilst CMS treats this as a single type A nuisance. It’s not discussed in detail but it seems here a similar thing can be done as for pT(Hjj), where we assume the main source of the CMS type A uncertainty is this boundary and thus it can be correlated with the ATLAS nuisance. Again this could be properly discussed in a follow-up meeting.