Cross section from xseccomputer is different from other papers or Madgraph

[wk-mike]

Hi，
Thank you very much for the useful software. I have used SModelS for a long time. It's great!

When I doing research about searching smuon, I find some problems about using 'xseccomputer' to calculating cross section.

xsection from paper:

According to paper arxiv:1908.08215 ( ATLAS-SUSY-2018-32 ) Page 5

the cross-section for ˜ℓ ˜ℓ production, each with a mass of 500 GeV, is 0.47 ± 0.03 fb for each generation of left-handed sleptons and 0.18 ± 0.01 fb for each generation of right-handed sleptons.

and from paper 1310.2621 Page 26 ; and paper 1801.10357 Page 8-9 , we can see that:

For 500GeV smuon_L and smuon_R, at 13TeV pp collider, the LO cross section is :
pp -> smuon_L smuon_L 0.42fb
pp -> smuon_R smuon_R 0.16fb

xsection from xseccomputer(SmodelS):

And we can do a quick check, the files
to_sm2.txt
to_sm8.txt
containted NMSSM spectrum and decay information (calculated by NMSSMTools) .
In both files, the mass of smuon is about 500GeV.

1000013 5.09478014E+02 # ~mu_L
2000013 5.03261715E+02 # ~mu_R

1000013 5.01350629E+02 # ~mu_L
2000013 5.07864400E+02 # ~mu_R

Noted:
smuon_L 1000013 ;
smuon_R 2000013

To get the cross section (using xseccomputer in SModelS), we can run the code
./smodelsTools.py xseccomputer -s 8 13 -e 10000 -c 1 -p -f to_sm2.txt
in the latest smodels-2.1.1.

And we have :

XSECTION 1.30E+04 2212 2212 2 -1000013 1000013 # 10000 events, [pb], pythia8 for LO
0 0 0 0 0 0 2.51334400E-04 SModelSv2.1.1

which means the LO cross section ( pp -> smuon_L smuon_L ) is 0.25 fb, and (pp -> smuon_R smuon_R) is too small to print. And it's very different from the paper

pp -> smuon_L smuon_L 0.42fb
pp -> smuon_R smuon_R 0.16fb

And I have to mention that when I change the number of event,
./smodelsTools.py xseccomputer -s 8 13 -e 50000 -c 1 -p -f to_sm2.txt
./smodelsTools.py xseccomputer -s 8 13 -e 100000 -c 1 -p -f to_sm2.txt
the cross section ( pp -> smuon_L smuon_L ) and ( pp -> smuon_R smuon_R ) have different values. (I do delete the old xsection in to_sm2.txt file)

xsection from Madgraph:

And we have used Madgraph to check this cross section.
I'm using MG5_aMC_v2_9_6.

import model nmssm
generate p p > mul- mul+
launch

import model nmssm
generate p p > mur- mur+
launch

we replace the param_card.dat with to_sm2.txt , and keep the nevent 10000, 13TeV collider. we get

Run | Collider | Banner | Cross section (pb) | Events | Data | Output

run_01 | p p6500.0 x 6500.0 GeV | tag_1 | 0.000324 ± 6.5e-07 | 10000 | parton madevent | LHE

run_01 | p p6500.0 x 6500.0 GeV | tag_1 | 0.0001338 ± 2.8e-07 | 10000 | parton madevent | LHE

which means the LO cross section ( pp -> smuon_L smuon_L ) is 0.324 fb, and (pp -> smuon_R smuon_R) is 0.133 fb .
It's larger than the xsection from xseccomputer(SmodelS).
And it's close to the paper

pp -> smuon_L smuon_L 0.42fb
pp -> smuon_R smuon_R 0.16fb

The file 'to_sm8.txt' spectrum using as input，the same thing will happen.

I think that xseccomputer is using pythia8 to compute the cross section.
The reason might be that some error in the settings，or I am using NMSSM model, the program only support MSSM, or pythia cannot calculated all SUSY process cross section in high accuracy only run once....

I want to know what went wrong.
Hope to hear from you soon!
Thank you very much!

Mike