Dear Marylin, Willi, Bill and Brian
Many thanks for your answers. 
In fact the D labeling is not 100% and with derivatization there are some H added to the mixture. So the mass 3 is significant and is directly proportional to quantity injected (standard curve over 6 concentrations linear at R2 0.99), even if not representing a 100% of the molecule.  Moreover the final analysis will be done on human muscle lipid extracts in which the D31 is really at low concentration in a complex lipid matrix. 
The problem Isabelle faces is that the peak she which obtain is not sharp any more and is tailing. It is tailing only for D31 and for non of the other fatty acids of the mixture. 
Few weeks ago the peak and shape were perfect. We have changed injector, cut the column, change the pyrolysis oven, the syringe. We are stuck.

Regards
Isabelle