CBS News was slated by supporters of Glenn Beck for its estimate of his rally attendance (87,000), so they published the methodology and here's their estimate for the Rally for Sanity (215,000). Both have a 10% margin of error. Their estimate for the Obama inauguration was 800,000 by the way, lower than others.
- Blog by the expert partnering the survey company. It has plenty of links and information about the subject.
- Plenty of first hand accounts of the latter rally.
- Overhead view of the main crowd areas - in both cases more people were further away:
When I looked up rational estimates by supporters, one reason for disparity was obvious - crowd density. In camera shots taken from the ground (or even the top of monuments) density isn't clear, but the company CBS hired used a robot balloon camera that could do both overhead shots and oblique ones to see beneath trees. It was on the spot and could tour the whole crowd, later estimators were not.
The conclusion is much as usual. How was the data obtained? What is the margin of error? Why do you suppose you can understand any of this without having done Statistics 101 or switched your brain on before reading?
Also, take a look at who is giving you different data with no sources and no proof. If they claim that a source is wrong because they don't like what it says and then next week that the same source is totally accurate, alarm bells should be ringing in your head.