Why doesn't madgraph5 give the correct ttbar xsec out of the box

I'm running MG5_aMC_v2_3_3 out of the box without pythia,pgs,delphes etc, just parton level, and when I generate the process:

generate p p > t t~

I get a cross-section of 505pb. The relevant parameters in the run card are set to LHC-like default conditions (13 TeV pp beam) and I didn't alter any of the other default settings. I've seen in tutorials online that this number of 500pb is expected, but I'm wondering why it isn't closer the NNLO value of ~830pb or the NLO value of ~900pb.

Is this a problem (perhaps I'm missing out some diagrams?) and, if not, how can I get a cross-section closer to the predicted values?