SUSY production cross section discrepancy between Prospino and Madgraph

I'm finding a very large discrepancies between the production cross sections for squarks using Madgraph (LO) from Prospino (NLO). While I'd expect factor of 2-3 deviations (roughly the K factors, possible efficiencies, etc.), I'm finding deviations by a factor of a 100.

I'm generating my cross sections using all the default settings (variable renormalization scale, no additional cuts, etc.) and set all the mixing angles to unity (in particular RRd1x1 = 1) in the mssm model. I'm simply using

> generate p p > sd1 sd1~

If I do this with msd1 = 1000 GeV using 4000 GeV / beam, I get a cross section of 0.0002689 +- 3.367e-07 pb. This is about a 100 times smaller then the cross section quoted by prospino in for example this paper, http://arxiv.org/abs/1412.2784, 0.02 pb.

I am likely doing something wrong but I have no idea what it could be. Any help on this would be greatly appreciated.