Relatively few things in nature are blue, or white. These colors are most prominent in the sky, during enjoyable conditions. So, the model, knowing reality through images, thinks in luminance (channel 0) cyan/red (channel 1) and lime/medium purple (channel 2), where Red and Green are primary and blue is secondary. This is why very often, SDXL generations are biased towards yellow (red + green).

strangely i prefer, in almost all of the examples, the nonColorCorrected yellowish images 🤷

thanks for posting. knowing this might help with future generations