I’m looking at the specs, and it doesn’t seem to specify whether incoming values are scaled first, then biased, or the other way around. I’m guessing it’s the former, but it would be nice to have this stated in the specs. Or if it already is, then it should be more visible, because I couldn’t find it. 
Andras
Without looking I think it’s multiply then add.
Edit: I came across the section you were looking for on page 138, 2nd paragraph. Unfortunately a keyword search for PixelTransfer won’t get you there, but it’ll put you in the ball park (or maybe just in the parking lot)
I checked both 1.4 and 2.0 freely available as PDF specifications.
3.6.5 of both the 1.4 and 2.0 spec spells it out:
“Each component is multiplied by an appropriate signed scale factor: … Then the result is added to the appropriate signed bias: …”
What could possibly need some work is making this reference obvious earlier.
Andras, what spec’s did you read not saying this?