I have found this to be true on Agilent and Varian GCs, as well as with process control GCs.
Improvements of detection limits ranged generally, from 30% to 50%, depending upon the analyte in question.
My understanding of the FID is that only a tiny fraction of the analyte molecules are ionized in the best of hydrogen flames and electrodes. Is it not likely that helium makes a better insulator of any possible tiny current than nitrogen?
Fortunately, we can use the improved results without having a good understanding of it, although said understanding might help us to increase the effect and improve performance.
Happy research,
Rod