Ok here goes
you get that additively green and red= yellow, g+b= cyan B+R = magenta?

cyan dye absorbs red
yellow dye absorbs blue
magenta dye absorbs green

add Y an M dyes and you get red
ye= -B
M= -G
so RGB -B -G =R

Y+C
C=-R

RGB-B-R =G

C+M
RGB -R -G= B

www.optique-ingenieur.org/en/courses/OPI_ang_M07_C02/co/Contenu_05.html

might be clearer visually. My text that this was quickly taken from had wavelengh graphs and lots more from a course unit, forgive the scruffy.

It isn`t as simple as all that google trichromatic matching, ( tristimulus values, colorimetry matrices to get some idea.

Edit: there is a basic problem which is that the colour receptors in the eye have ovelapping irregular responses. in some cases given three suitable illuminants you can match them to tickle the eyes receptor exactly to the same value of a particular colour and brightness.

However It all gets a bit difficult. Monochromatic single wavelength excitation in cyan at about 480 nm will produce responses in the blue and green eye receptors and virtually nothing in the red. If we try to match that with a blue and green illuminant at at about 460 and 500 to match the eye`s peak response the green primarry stimulates the red receptor so that the colour looks too red. If using a visual matching colorimeter you need to add red to the primary cyan to match the two RG primaries. or to put it another way you have to have a -Red stimulation....WTF? In Tv systems you can matrix the sensor primaries so that the crosstalk between eye sensors is compensated by subtraction between the colour channels.

Typical daylight illuminance still leaves a few colours that can`t be resolved in the PAL or NTSC colour spaces for phosphors. This is a huge problem for LCD displays which is slowly being better addressed with LED back lighting, but there you go. the backlight should be a proper tristimulus tube like a scanner one and the transmissive dyes matched No idea how much more restricted the chromaticity triangle of subtractive dyes is.
(tootired Is searching for a funny sig, nothing good so far, Tue 18 Oct 2011, 12:48, Reply)