Hi Daniel,
Sorry I took a long time before being back to you.
Thanks a lot for your advice. I was indeed able to follow the examples 5a and 5b of the afni_proc.py. I have now a script that works fine and everything seems to be well taken into account.

I used RetroTS.m.

But I'm remaining with my first question :

To use RetroTS.m we have first to chop off unneccessary values in physio files by aligning the timings and removing evertyhing before the beginning of the scan and everything after the duration of the scan. (I also removed the 5000 and 5003 values added by Siemens)
To do this I used the "LogStartMDHTime" (an not the LogStartMPCUTime) that can be found also in these raw physio files. And I aligned this value with the "AcquisitionTime" that can be found in the DICOM header of my fMRI files.

Is it the correct method?

Thanks a lot for your help.
Alain