I think a possible source of confusion here is that colour is continuous not discrete - it's convenient to talk about "red light" as though it had a single definition, but in reality, "visible light" can be composed of any combination of the entire spectrum of wavelengths that our eyes are equipped to detect.
So when you say:
... each pixel ultimately captures one of red, green or blue colors of light ...
You are missing that our colour vision works not by picking three points on the spectrum of wavelengths, but by comparing three overlapping sections of it. We can trick the eye by displaying three discrete wavelengths, which match the peaks of these categories; but to capture a scene, we need to record what the eye would see looking at it. So even on a digital camera with an explicit "RGB" filter, there are a wide range of wavelengths being captured.
Similarly, when you say:
... red, green and blue colors of light bend differently ...
You are missing that there are not three choices of "bendiness" that light can choose, but a continuous function where any wavelength you pick has its own distinct focal length. So any scene projected through a lens has a whole array of different wavelengths, all focusing slightly differently.
Put those two things together, and you'll see that the filter in front of a digital camera sensor can't correct for chromatic aberration, because the "red" sub-pixels still need to capture both the "reddish-yellow" light and the "yellowish-red" light, which are focussed slightly differently.