I don't think they use LED's, since you'd only get a small handful of discrete wavelengths to play with. Hence they use an incandescent light and a monochromator (variably positioned prism and a slit) to give you your light. (Unless you've rigged up your own kit in which case LED's would work in most cases, just means you only get one type of blue.)
If you use blue light with blue solution, all light incident on the sample will be reflected. The colour gizmo compares the light input to the output, so it knows how much goes in, and how much comes out. The difference is the amount your blue solution gobbled up, ie. the amount absorbed.
The K_3Fe(CN)_6 will certainly affect the reading of absorbance (I'm guessing atleast). However you find the absorbance of a set of known standards, then you prepare a calibration curve. From this you find the concentration of your sample by seeing what its absorbance will be.
So your sample's solution must be within the range of your standards, if need be dilute your sample and then bump up your answer later (Similar to volumetric analysis techniques).