Different numbers of RGB % between the eyedropper and the histogram

The WB loupe readout averages the values, but not sure if this is center-weighted or evenly taken from all the pixels in the square. Regardless, the histogram readout operates the same all the time including when in WB, and in other contexts does not present any sampling square. I suspect the histogram readout is either constantly sampling a single original pixel at the cursor coordinates, or else (IMO more likely) a single screen pixel at the current zoom, but either way is disregarding whether the WB loupe sampler might be set large or small, and what other nearby pixels it may or may not be therefore including into its own average.