Thanks to the both of you!
Just to make sure I understand correctly: Does this mean, the 3x3 reading is from 3x3 pixels of the actual preview screen-pixels (viewer), thus depending on the size of the preview? I just did a quick test and it seems it could be that way, but I have to check it out. As you cannot make the viewer show the preview smaller than "fit in window" I can just make the whole window smaller to get a smaller preview - would be at least a way to work around my "issue".
That would do it i guess.
(Still wondering if I'm the only one with it. It would be nice to have the ability to e.g. just somehow "select" the whole grey-card for WB, preferably in any size, a mean would always be more exact as a few pixel measurement, wouldn't it?)
There may be uses for it, but white balancing is quite complex if you think about it. It's not just about getting equal R/G/B readouts, although that's how it's usually used, but it is trying to match the color of the light-source with a presumed spectrally uniform reflecting surface. However, other than a single light-source in a dark room that illuminates an isolated subject coated with a special quality white or gray, there is no such simple situation in practice.
We have to deal with an imperfect surface reflection under mixed lighting conditions with ambient reflection from surrounding subjects/surfaces. This is also assuming there is little noise, because noise usually changes the color of Blackpoint demosaicing. The best one can do is get an estimate of what the predominant light source's emission looked like, and then mix that information with the Profile assumptions for a given colorspace.
So the question then becomes more, what is the problem we're trying to solve, rather than how to get R=G=B.
Maybe there's something that can be done by using controlled lighting, or maybe adding a low ISO shot just for white balancing will do better, or use noise reduction if we couldn't increase exposure at capture time, or average multiple shots (something that I'd like to see in future versions of C1) to improve the Signal to Noise ratio, which should allow easier white balancing.
Just some considerations.
Cheers,
Bart