Understanding the cross-section of Drell-Yan of whizard vs MG

Hello!

I am using a mili-charged dark pion model (UFO) in whizard. The model is used in conjunction to SM, so the Drell-Yan process is:

proton ,proton => dpip, dpim

Where dpip and dpim are milicharged dark pions. This process usually has the virtual photon and Z as propagators, but in my model, we also have a dark Z included. I compared the result with another event generator (Madgraph).

Here is the SINARDIN script:
#############################
model = pidm_UFO_v1 (ufo)
alias proton = u:U:d:D:s:S
process testdy = proton ,proton => dpip, dpim
beams = p,p => lhapdf, lhapdf
$lhapdf_file = "NNPDF23_nlo_as_0119"
?rebuild_library = true
sqrts = 14 TeV
n_events = 1000
$phs_method = "wood"
?write_raw = false
?rebuild_events = true
scan Mdpim=(1 GeV => 100 GeV /+ 1 GeV){
    MD=MU
    MS=MU
    kappadark = 0.01 GeV
    fdecay = 100 GeV
    compile
    $run_id = sprintf "testdy100_%1.3f" (Mdpim)
    $sample = sprintf "testdy100_%1.3f" (Mdpim)
    sample_format = lhef
    simulate(testdy)
    show(results)
}
#############################

fdecay is the dark pion decay constant. kappadark is the mixing between SM and DS. I set the width of dark Z to be 9.914352e-06 GeV to ensure both event generators have the same width. The PDF for the process is the standard nnlo23 PDF (pdlabel = 230000). The mass of dark Z is set to be half the mass of Z (MZ' = 45 GeV). I had to set the masses of the quarks to be equal because otherwise alias fails.

I'm not sure if it is possible to attach an image so I uploaded it to imgur: https://i.imgur.com/ACsd7L0.png

As you can see, MG shows a steep drop in cross-section once it crosses half of MZ' (~22 GeV), we do not see that in WHIZARD. This can be explained by the on-shell contribution by the dark Z before resonance. Why do we not see that in WHIZARD?

Am I missing something or not including something in the SINDARIN script?

Thank you.