SMEFTatNLO event generation

Dear Madgraph Experts,

     We are generating samples to parameterise the contribution of EFT operators in SMEFTatNLO for gluon-gluon fusion Higgs production. A peculiar behaviour regarding the usage of restrict cards for particular values of the coefficients was observed and we would be glad if you could throw some light on the matter.

     Considering the generation command template below, where restrict cards are loaded for coefficients mentioned in the restrict are set to 1.0 and the rest are set to 0.

     import model SMEFTatNLO_U2_2_U3_3_cG_4F_LO_UFO-NLO_{restrict}
     generate p p > h [noborn=QCD] QED = 1 NP<=2 NP^2=={val}

  We see the following behaviour,
        When {restrict} is c3pl1 and c3pl2 the predictions are equivalent and opposite to cll1221 for val=2.
        When turning on two operators at a time (c3pl1_cll1221, c3pl2_cll1221), we see that if c3pl1 = cll1221 or c3pl2 = cll1221 in the restrict card we get a No Diagram Exception for val=2 and val=4.

   Furthermore, the prediction for NP^2==0 is the same as NP^2<=2 and NP^2<=4 which suggests that for c3pl1 = cll1221 (a) or c3pl2 = cll1221 (b), the contribution is the SM prediction. This behaviour also appears for the case c3pl1 = -c3pl2 (c). For any other parameter values that don't satisfy the previous equalities, we seem to avoid the no diagram exception.

   Is the No Diagram Exception for such a scenario a nominal behaviour ? Does the cancellation of the EFT contribution for NP^2==2 when any of (a) (b) (c) is true necessarily guarantee that the contribution at NP^2==4 is zero ? Thanks a lot for your insights.

Cheers,
Ana, Saskia, and Rahul