ttbar + 0,1,2j discrepancy between madgraph versions

Hello Jon,

The reason for the difference in cross section is a change that was done
in v. 1.3.18:

1.3.18 (04/10/11) ...
                 JA: Enable proper 4-flavor matching (such as gg>hbb~+jets)
                     Note that you need the Pythia/PGS package v. 2.1.9 or
                     later to use with 4F matching.
                     Changes include: alpha_s reweighting also for b vertices,
                     new scale treatment (mu_F for pp>hbb~ is (pT_b^max*m_Th)),
                     no clustering of gluons to final-state massive particles
                     in MadEvent.

Since you are running with maxjetflavor = 4, it is considered 4-flavor
matching, and therefore there is alpha_s reweighting of the bbg vertices
(where the final-state b radiates a gluon).

Note however that this process will *NOT* get correctly matched.
Matching of QCD radiation off decay legs (such as the b from t->Wb
decay) is not treated correctly. For the MadGraph-Pythia matching to
work, you have to restrict QCD radiation to only the core process, i.e.,
you should use the following process definition:

generate p p > t t~, t > b e+ ve, t~ > b~ e- ve~ @1
add process p p > t t~ j, t > b e+ ve, t~ > b~ e- ve~ @2
add process p p > t t~ j j, t > b e+ ve, t~ > b~ e- ve~ @3

In fact this will still give a very good description of the gluon
radiation from the top decay products, since Pythia has built-in matrix
element corrections for the parton shower off top decays.

All the best,
Johan

I answered the question above.