There is a lot of confusion surrounding UniWB.
1. Most of the attention given to it concerns how to set it up which can be easy or difficult depending on who describes it.
2. Very little attention is given to what it actually does and how to use it if you get it set up.
3. Finally, almost no attention at all is given to whether it actually accomplishes anything useful.
I found a simple way to set my Nikon Df to Daylight and UniWB. I then captured a scene based using each WB setting. The scene was chosen because it contains an extremely reflective white PVC fence, brighter than any sunlit cloud or white bird.
Both images were exposed at 1/250s f/11 @ ISO 100, 1/3 stop darker than Sunny 16. Although the camera was not mounted on a tripod, the exposure and ISO were identical in each case so both raw files are virtually identical.
In fact the two luminance histograms on the camera were very similar since they are based primarily on the green channel. However, there is a slight difference in the far right end of the histograms:
Looking at only the luminance histogram you might be tempted to increase the exposure a little. But as you will see, this might be a mistake.
What can we learn from the RGB histograms?
The main difference is that in the UniWB histograms the red and blue channels are pushed significantly to the left leaving the green and luminance channels where they were - almost.
Since the two raw histogram displays are identical except for the associated image we need only one of them:
You can see from the raw histograms that it might be possible to add about 1/3 stop to the exposure and actually use Sunny 16 - 1/200s f/11 ISO 100. But that's all. Any more and the green channel will blow out.
But if you look at the camera's histograms, the Daylight WB version already told us that. UniWB didn't actually tell us anything significantly different from what we could see in the Daylyght WB version in the camera's display.
There was no way that we could expose any than maybe 1/3 stop further to the right and that would not make the raw data visibly or measurably better.
So the bottom line is that there is no demonstrable benefit to UniWB.