Color theory, or atleast its naive implementation, is almost certainly somehow broken. Why?
If you are selecting based on a color wheel. Then your entire selection crieria is perdicated on the fact that the color wheel is correct. But you have many different color wheels. Thus many different selections are possible to fullfill the result it just depends on what colorwheel you use.
Color science answers your question here better (as opposed to color theory). It turns out that we can measure what the diametrically oppsite color of each color is. Because we can measure the after glow of the color. We can also see how the sensors are laid out on the retina. So we have concluded that the oppiste of red is green and the oppsite of blue is yellow.
Color science also tells us that there is a fancy white balance on top of the system. So you should be able to put a color filter on top of the system and it should work just as well because the whitebalance shoukd handle this. OK so if you now put a red filter with a hint of orange on top you should be able to move them into a position where they are allmost dimetrically opposed colors.
This nicely explains why the color wheel construction matters as little as it does. But also explains that the color theory aspect probably has no super deep meaning after all.
The exact colors problably jitter a bit because they were almost certainly selected form some color mixing guide book.