Good morning,
mea culpa! The problem vanished.
I had forgotten that SilverFast first applies the gamma correction and then
the ICC-Profile. Since I had derived the ICC profile from the linear data,
the gamma correction in SilverFast got applied twice. Thus the shift to
brighter values. One has to derive the ICC profile in Argyll from the
gamma-corrected data.
Sorry for the disturbance.
Hermann-Josef