As far as I can tell from the conclusion, everything was within the margin of error, so my assumption is that it's random noise. If there's a place in the paper that says otherwise, please let me know what page its on. If I'm misreading the results, please let me know that as well.
The 40% figure seems to come from section 8.2, p.152, which the authors describe as "robust".
However, having seen the report now, this section on top films seems to use a different methodology to that used for books, so it's not really comparable, and in general I wouldn't put much confidence in these results anyway.
