I have been aware on a quite unexpected behaviour of j_src_spectra, and I would like to ask you if this is normal.
The net count rate given by XSPEC for a source spectrum depends on the the energy channels chosen when running jemx_science_analysis. I thought these channels only affected imaging and light curves. However, I have observed a difference up to about 20% in the spectrum net count rate only by changing the energy channels.
This effect is the most obvious when using short GTI intervals.
I send you as an exemple two spectra and the corresponding ARF (the two are identical) together with their respective log files and XSPEC log.

What do you think?

Cheers,
               Jérôme
 

-- 
Dr. Jérôme CHENEVEZ                     Danmarks Rumcenter
Danish National Space Center            Juliane Maries Vej 30
http://www.dsri.dk/~jerome              DK-2100 Copenhagen Ø
Tel. +45 35 32 57 03                    FAX: +45 35 36 24 75