We don't know the exact makeup of our supporters. If you go to
http://rally.ronpaulplanet.org/stats.php you can see the stats of those who have posted pictures.
Now, you have to understand that this is a self-selected sample so it's incredibly prone to bias (if you don't believe that, check out this digg article
http://www.digg.com/2008_us_elections/Survey_Majority_of_Americans_Agree_with_Dennis_Kucinich - according to this self-selected sample of 60,000 Americans, we're all socialists).
That said, I think the numbers shown from the rally site are probably in the ballpark at this point. About 25% traditional Republican, 30% independent, 20% Libertarian, 12% Democrat...etc...
Now, I don't know the polling methodology for the various national and state polls, but in likelihood they are applying sound statistical practices to predict primary races. Understand, they've been doing these polls for many years and are able to use past data to find the best methodology of replicating the intended population.
What you need to realize is what population they are trying to sample. It's not the general public, it's not people who occasionally vote in primaries - it's essentially consistent primary voters who lean Republican.
I think nationally, about 25-30% of registered voters show up for the primaries (someone correct me if i'm wrong). Only half of those are going to be republican primary voters. That's the "target" population they are trying to quantify. It accounts for about 15% of registered voters. If you aren't in that group, then you aren't part of the target sample population.
BTW, statistically, this is totally legitimate. They have years of primary polling experience that tells them, year after year, this is the group that actually shows up. If there's is a huge error here, completely overlooking Paul support, it's in the target population. In other words, they ARE measuring their INTENDED population correctly - the flaw will be that the "old" target population of previous primaries no longer matches the "current" or "real" population.
IF either party had a candidate who was able to draw significant independent and 3rd party support for the primaries, I don't think they'd know how to account for that.
Until someone (grassroots project maybe?) commissions their own poll to try and figure out these discrepencies and understand these new dynamics, it's really hard to guess what an actual primary vote would look like.
The above is just my opinion based on a couple of stats classes in college...