Crosssections for uu~>tt~ scattering

Dear MadTeams,

I have a question on the cross section calculated by MG5.

In the following procedure, we can obtain uu~ > tt~ scattering in "Tevatron-like" setup,
and cross section is calculated:
| ./bin/mg5
| > generate u u~ > t t~ QED=2 QCD=2
| > output UUTT
| > launch UUTT
| and here edit "run_cards.dat" as: lpp1=0, lpp2=0, beam1=980, beam2=980.

In MG4, the cross section was 1.89pb, while MG5 gives 1.66pb.

I doubt this result and I tried another choice: I did the same procedure as above, but with QED=0 QCD=2.
Here MG4 gives 1.87pb, and MG5 is 1.87pb.

These results seems that MG5 does some weird things with multiple diagrams.

I am sorry that I could not find the bug, but I have two clues:
(1)
In QED=2 QCD=2 case, we have 3 diagrams: s-channel with gluon, photon, and Z.
but the calculation gives only G1 and G2; there is no G3.
(2)
In QED=0 QCD=2 case, the generated "configs.inc" has
      DATA SPROP(-1,1)/21/
(21 is gluon PDG!)
and QED=2 QCD=2 case, it is
      DATA SPROP(-1,1)/22/
      DATA SPROP(-1,2)/23/
that is, only photon and Z; here no gluon.
I doubt this file, and modified "22" to "21" by hand.
Then the cross section is calculated as 1.88pb, and the result seems to get better.

So now I think that
  * one diagram (glu-exchange) lacks, or
  * configs.inc is badly generated

I am sorry that I did not investigate the source code.
I am glad if you could give some ideas.

Best Regards,