Thank you very much @chris,
the glass is entered as BSDF, calculated with LBNL Window, I will leave you the datasheets here.
I think it only takes the wrong values as much as it is at high incidence angles.
I can turn the .idf file over to you privately.
Thanks again for your support!
Massimiliano
UPDATE: Sorry @chris , I got confused, for the comfort map test I used simplegalzing and put 0.999 to the inner dummy glass! And now it works!
If it is of interest I had simplified the model, removing the construction with the bsdf and converting the schedule:compact and schedule:constant to day/month/year split schedules because the HB Load gbXML IDF component did not read the .idf file
