I strongly recommend using RAW images. jpg are not useful for science.
How are you measuring the pixel value in the image? Is this the value in a monochrome image made from the colour image? If so, check that none of the RGB channels in the colour image is saturated. (for example the G channel could be saturated but not the B and R channels and you would no know this from the mono image) This could also give the non linear effect you are seeing.
Cheers
Robin