I'm glad others have noticed that this forum poll type is different than the previous blog poll type. Here's my take:
The 29% that Cinnamon has on the forum poll does *not* mean 29% of the people voting are using Cinnamon. Apparently the "total votes" number represents the total number of selected DE options across all persons participating in the poll. This is evident in the fact that multiple options are allowed to be chosen, yet the percentages add up to 100%. (This 100% can vary slightly due to rounding.) This poll roughly shows the percentage of installations of a DE, but only roughly, as one person could have multiple installations of the same DE.
This is in contrast to the blog poll where the percentages added up to well over 100%. At the time I recorded the numbers, it was 150%. Here the percentages were actually based on the total number of persons participating. So the 41% for Cinnamon means that 41% of people voting use Cinnamon. But with the poll alone, we cannot jump to the conclusion that Cinnamon is used more often than the other DEs, just used by more people. (It could be used infrequently by lots of people.)
But that's far from the whole story.
No poll is perfect: The blog poll had more non-Mint users (no doubt), the forum probably more Mint users. (Although, from the forum posts, it looks like many people registered on the forum to vote.) Did the blog poll have more Cinnamon users who are not Mint users? How many participants answered for others? (I see many posts/comments from people who install an OS for other people.) There are many such questions and issues.
The real questions are, How often is a DE used? How useful is a DE compared to another? Unfortunately, neither poll can answer these questions definitively. I doubt any poll could.
Consider this example:
Someone votes for MATE and Cinnamon. However, MATE is installed once and Cinnamon 5 times.
1) Perhaps all installs are used equally so Cinnamon is used 5 times as often as MATE.
2) Perhaps MATE is on a person's main machine and the other Cinnamon installs are used very infrequently. Here MATE is used a lot more than Cinnamon.
3) MATE is used Monday to Friday; each Cinnamon is used only once a week. Here they are used about equally.
So it seems that multiple sources of information need to be analyzed to get the best answers to the real questions. Below we compare the Forum Poll % v. Blog Poll %.
After that we see that the Progression of Forum Votes is flat for each DE: This poll had the same percentages for each DE whether there were 555 votes, 774 votes or 2027 votes.
We can infer (from this and many other polls/votes) that usually a small sample is representative of a much larger set. Depending on several factors, this is not always the case. But I think it's safe to say that the blog numbers (shown below, before poll corruption) are representative of what that poll was asking.
EDIT: Fixed link to image. Thanks zerozero.