I have now read your (Andrew's) article on white point and re-read the ICC white paper.
I am aware of that scene-referred data have to be translated (rendered) to human vision, which is non-linear, unlike the camera sensor. This process of translation will necessarily imply individual judgement, in particular w.r.t. the tone curve. This is in my eyes NOT a carte blanche to whatever arbitrariness. We should try to apply a terminology that at least theoretically enables us to describe the difference between a naturalistic, if by necessity subjective, rendering intent and e.g. an intended change of hues. Using the term "pleasing" for both is confusing.
I am painfully aware of, that the in-camera jpeg is only very loosely linked to the raw data - because the camera maker does exactly what you advice me to do: just make it look pretty. I am not exspecting or claiming that my "method" gives color accuracy to the 3rd post comma digit. Nevertheless, with a precision level as coarse as 1,000°K, it could produce an image, that was MUCH closer to the scene than the AWB. Just picking the prettiest might have led me to #3.
The bracketing is only intended for the visual comparison, not believing that it will change the raw data. I will trash the shots after I have decided on the WB.
The magenta-green axis
I am aware of that that this is to be considered. But how much does the daylight change during the day along this axis? In another thread on this forum, which I fail to retrieve right now, I was told that it changes very little, and have since set the Tint in the raw converter to zero.
The difference between different (well, 2 so far) raw converters seems to be far less than the differences between the AWB and my "method".
Let's turn it around: Which are my alternatives?
AWB - see above.
Gray card, WhiBal and the like: Will theoretically make every image look like it was shot at noon.
Using a fixed value of say 5,500 K (which is what I have done lately): Well, #2 of my images is 6,000 K, and it's way off.
Maybe I can put it this way: the in-camera jpeg, despite all its shortcomings, still seems to be a tool that is my best bet in this situation. Better than AWB, and better than my memory after the shooting.
Good light - and true color
Bill, this was written before I read your post. I may return later.