There are at least two reasons this happens:

1. There are very few pure (single / narrow frequency band) light sources. That is to say, your red light is not strictly confined to the "red" end of the visible light spectrum. It emits light in the yellow/green region of the spectrum (and probably just a little in the blue, as well). Similarly, your blue light is not confined to the blue region of the visible light spectrum.

   Lab-grade lasers emit very narrow light spectra. They are probably the closest "single pure color" light source you can use. Most other light sources have much broader emission spectra.

2. There is substantial overlap in the light frequency sensitivities of the color filtered photosites in your CMOS sensor. That is, the red-sensing sensor pixels (more appropriately, "reddish"-sensing) overlap quite a bit into the greenish spectrum. The blueish-sensing pixels also respond a little bit to greenish and reddish light, as well.

   There is nothing you can do about this. Tri-color imaging sensors are made to mimic the human eye's response to light. The "average" or "normal" human eye color receptors, while described as red, green, and blue, have a lot of overlap in light frequency. This is especially so in the red and green receptors, to the point that the most common form of color deficiency or colorblindness is red-green. This is because the frequency response of the red cones and green cones in red-green deficient or colorblind people overlaps so much, that they can't distinguish reddish hues from greenish hues.

If you want more color separation, you will have to get better / more pure light color sources. But even if you use a single-frequency light-blue light source and a single-frequency red light source, they will probably be picked up in the green sensors to some degree.