Correct me if I'm wrong but, since Blurb is the commercial name for a number of different labs with different printers which all respect a sort of "standard", what's really the point in trying to perfectly profile the specific output of a specific (yet unknown) lab?
I can't answer that. If I were to send the books to the company many times over a span of weeks or months, I could get a really good idea how close each press and the overall process control is (within reason from someone trying to do this from the outside. From the inside, I could gather very exacting info).
I can tell you this: I had one book printed from Lightroom just after version 4 was released. It was pretty awful. I sent another book (same images) a couple weeks ago, the output was much better! So I don't have a lot of faith in Blurb's QC and process control but keep in mind, we're talking about two samples over a course of about a year.
Lastly, I'm not trying to make a profile to do anything but soft proof! The target inserted in the book was RGB such that it would move through the entire process (from LR, to whatever RIP or front end Blurb uses from LR, through the press). My idea was to build an RGB profile only
for soft proofing in LR. I can't convert to CMYK for Blurb nor would I until I had control over what gets sent, in CMYK to lots and lots of presses, with lots of different targets after seeing how far the dE values vary for each press and after averaging the 'good' data. Then I'd monitor the CMYK process by sending the targets again to these presses over time and comparing those measurements against what the profile predicts (the reference).