Quad steel/conc stresses NONL analysis

Hi all.
I’m trying to obtain stresses for fatigue veryfication - quad element under bending.
I made two separate load cases 7001 for my permanent loads and 7002 for permanent + cyclic loads.

I have a sheet for manual veryfication which gives me similiar results to stresses values directly from WinGraf - of course when I put the same values of reinforcement and bending moments.

And there is my problem..
When I first calculate these 7001 and 7002 LCs I get values let’s say M1 and M2. Then after (presented below) ASE run I get these M1 and M2 values rewrited with much higher values e.q. M2’ almost 2*M2.

Here is my ASE code:

+PROG ASE urs:49 $ Non-linear analysis ULS (Area elements)HEAD Non-linear analysis ULS (Area elements)
PAGE UNII 0
CTRL ITER 3
SYST PROB NONL ITER 200 TOL 0.002 NMAT YES

NSTR KSV SL

REIQ LCR 1 FACT 1.0 LCRS 98 $ use reinforcement from design case 1, new reinforcement saving in design case 98

LC 7001

END

+PROG ASE urs:51.2 $ Non-linear analysis ULS (Area elements)HEAD Non-linear analysis ULS (Area elements)
PAGE UNII 0
CTRL ITER 3
SYST PROB NONL ITER 200 TOL 0.002 NMAT YES

NSTR KSV SL

REIQ LCR 1 FACT 1.0 LCRS 99 $ use reinforcement from design case 1, new reinforcement saving in design case 99

LC 7002

END

Sorry for my ignorance, I know that after nonlinear analysis higher forces may occur, but could the difference be such significant?
I assume the correct approach would be to use bending moment values after NONL analysis for the manual check?

Best regards,
Szymon

Probably under perm loads concrete is uncracked. You have to reduce tensile strengh of concrete or perform additional unloading loadcase to obtain real stress difference