The FAQ shows some of the guesses that NewsBlur takes to get data out of blogs for better displaying the posts like they were in their original websites.
The schema.org website proposes a standard way to describe content through the use of HTML5-compatible microformats, providing basically _all_ the data that NewsBlur would need, without guessing.
It would be nice if instead of trying to guess where the data is on a blog, it would first check whether schema.org data is present on the page already.