I understand the difficulty of dealing with anonymous edits, because many of them might be edits from registered users who simply did not bother to log on for that one edit.
However, I think it is worth looking at how the conclusions might be affected under different scenarios for labelling those anonymous users.
For example, one might assume that the bulk of anonymous edits are made by infrequent contributors who are part of the long tail, as opposed to the members of the core. Does that change anything to the conclusion that most of the value is produced by a small core? If the answer is that even this does not change the conclusions, then case is closed. But if turns out that the conclusion is sensitive to how you label anonymous, then it seems to me that the next research that needs to be carried out, is to try and characterise the degree to which anons are, or are not registered users who are part of the core.
Alain