Circular systems like this are why Google applies a damping factor of approximately 0.85 to each transfer of link juice. If it left the juice at 100%, then the PR would exponentially increase until the central hub became the largest hub in the network, a humerous if undesirable result.
Now, my theory on why linkwheels work is not reliant very much on the PR system. This is because web 2.0 properties all come with a certain number of forced links which you cannot remove. As all PR is distributed equally out of a node, the more forced links you have, the less linkjuice you can direct to your money site.
I have compiled a list of web 2.0 properties, and calculated the % linkjuice which can be directed at a money site from each with a certain number of links. I have taken internal links that lead back to the porperty into account. Here is the table:
As you can see, you can use web 2.0s for link juice transfer, but they are generally not very efficient at it.
Yet, linkwheels work amazingly well. I think this is due to 2 factors.
1) Your link already has a reasonable amount of linkjuice from the web 2.0 service.
2) Your link is in the main text body, rather than in a comment. It is surrounded by text, and is on a site which Google recognises as relevant to your money site. It is generally accepted that Google's algorithm can automatically detect relevance or a lack thereof, and score a link up or down accordingly.
Regarding the diagram presented above, I feel that a) it is too simplistic in its link structure and b) 6 properties is far too few. It's a good start and a great concept but there really is no necessity to close the wheel.
In terms of the classic PR system, web 2.0 properties are just not that effective in transferring raw link power. Their strength lies in relevance, numbers and their own potential to rank due to Google's trust.
