The problem is that the user interface is broken! The white balance actually works correctly, the user interface is just impossible to understand for any sane human being. I have complained about the same thing in another thread.
The problem is that by default, if you uncheck the white balance, it will use the white balance values from the picture of which you had created the preset but that behavior is not visible to the user, as the user interface might actually show other values. The solution is to never ever turn the white balance module off. If you have any preset that has the white balance module turned off by default, it will just use any white balance, the behavior is not clear to understand in my view. Just turn the box on, and set the white balance either to ‘as shot’ or to a fixed value. ‘As shot’ will read the values from your camera. If these values differ, that is normal, every raw program interpretes white balance a bit different, they will be different in Lightroom, Capture One and Photolab. But the colors should be correct.
Does seem strange though that the “As Shot” is not, well, as shot. If it were it would not matter how many different settings the op used. Or am I missing something? And if it is not as shot why does it say it is!
When opening an image for the first time the WB is unchecked and the shown value is 5400. Selecting the WB does change the value but not the image. I still think the 5400 is an initial value used internal meaning no correction to be done. Checking and unchecking doesn’t change the WB temp anymore. I do remember an experiment of @Joanna shooting a graycard. The camera showed a histogram where the 3 channels covered each other more or less, but loaded in PL not.
I still plan to do that experiment myself.
That I know and isn’t the problem.
WB is always done, even if the WB isn’t checked. In fact, checking the WB means more or less that the original WB can be overwritten. When changing the color temp, the WB is automatically checked.
What’s wrong is that the initial value of 5400 is shown while another value is used and that that is only corrected by checking the WB.