Sorry,
I think I messed up a lot here, following some random results of rtrace runs. First, the results of rtrace illuminance with a sky dome defined as light are more or less random. So the fact that I got a result close to what I expected only led me to the wrong direction.
Defining glow, there is no need for multiplying anything for the colorimetric balance, as I assume a neutral white light source. The multiplication only has to be done when getting from radiance (which is in RGB channels) to illuminance. So I have either to use a glow with RGB 1 1 1 and modify it with the brightfunc from gensky (as is usually done) or directly use the (neutral white) radiance calculated dividing luminance by 179 for the glow material in case of an uniform sky (RGB 17.8 17.8 17.8).
Results are fine (9998.5 for the simple glow and 9987.6 for the skyfunc sky), I forget about my light-modified source, and hope that noone else spent time on my rather silly question...
CU Lars.