So, given the shift in wavelength given below, blanking isn't as useful with coomassie blue as in other situations. This is problematical, but not a disaster. You mention working with a 1 ml sample size, that should put the fluid level in the beam path, but do be sure. For example, a serial dilution of the Coomassie Blue, without any sample, should show a linear absorbance, even if the wavelength with maximum absorbance isn't the same with protein bound. You could try evaluating that, to conform the instrument is working correctly and you're preparing things correctly.