I find what 2F/2F says very convincing. Just to visualize it, imagine a portrait where the main light is a very lateral one (going by heart, Churchill's portrait by Karsh). Pointing the dome at the camera would bring overexposure. One does not want to "average" light and shadow, but to expose for light and then devise and use shadows to give more or less volume, "drama" etc. to the subject, so it is the main light that should drive the exposure.
I always use reflected metering but I'm going to get out with my incident light meter next time and do some practice. Maybe the dome of certain light meters is made in a way that it reads the same value both if the light source is at 0° (front) or at 45°. Maybe the dome of other incident light meters is more "directional" and that would lead to the problems pointed out by 2F/2F.
Besides, in a test in my balcony, my Gossen Sixtino II gave, in incident light, results that are way off what I get with the Gossen Multisix and what one gathers from experience. The Sixtino II in incident light, pointed toward the camera, would lead to gross overexposition. For many years my Sixtino II was all I had for incident light and I considered incident light measuring for years as a devilish trick or a mystery.
The Multisix has an hemispheric dome. The Sixtino II has a curious rolling shutter that must be moved in front of the cell. That I suppose is quite directional, and now I infer that the Sixtino II must be used the way 2F/2F points out, toward the light source, to give reliable results. I'll do some text when I have a proper sunny day.