the blue pixels would register the blue photons and the red pixels would register the red photons just like our eyes do.
No, sensels seperate wavelengths relatively sharp via filters, while L- cone cells are still somewhat sensitve to short wavelengths; akin to the spectral response of a Foveon sensor.
Take a sample of cobalt violet for example, light reflected of it as no spike in the red band, it absorbs red light about as good as black.
Technically true, the response curves of our cone cells do not have sharp cut-offs but please define "somewhat sensitive to short wavelengths"... if by that you mean "close to zero" then you are right. If you observe short wavelength light your L cones register a tiny response but the response in the S cone would be orders of magnitude higher and your brain would register that as blue light. Similarly incident light that is green would cause a response in all cones almost equally but your brain knows that is green, rather than white because of the relative responses to blue and red components.
Our brains have to be more complex to deal with the overlap and larger range of response patterns but that still does not mean our eyes, or a sensor, would be confused by UV light.... your eye simply will not see UV light as purple... our visual perception of "visible purple" is NOT the observation of near UV light.