I found it a very interesting paper and I caught myself nodding in agreement at the points that were made throughout.
I think separating the roles of facilitator and moderator in the interactive model would have addressed many of the concerns of the No.10 forum participants. I also think many of the causes for those concerns wouldn't have arisen in the first place with such an approach.
One further thing does occur to me - if the originals of any censored messages were preserved in the system for review by anyone who cared to look, then people would be able to come to their own view as to the balance and effectiveness of the moderators. That's essentially what you've actually done by analysing the quiscustodiet cache.
Moderators could tag messages that they removed from general view ('obscene', 'contravenes rule 6', 'off topic', 'spam', etc), possibly with a brief note attached where clarification was required.
Of course, then you get into the whole meta-moderation game.
Slashdot has the concept of moderation and meta-moderation carried out by the community itself, and has a sufficiently large community to mitigate against the Daily Me effect. It also embodies the concept of tagging messages in a positive way ('interesting', 'funny', 'insightful', etc). eg For me, reading Slashdot at 0 threshold is time consuming and annoying, so I usually read at threshold 4 - which cuts out the vast majority of the crap.
I really look forward to learning what the reaction to your paper is!