Well for one thing, most of the people who submitted to fueleconomy.gov also submitted to greenhybrid, so there's alot of overlap in the data.

For another, both databases are not very useful because they compare the driving of completely different people. The most useful and scientific data is from the same person driving different cars and comparing the mileage for each. I own a 2006 HCH and I've rented a 2005 Prius twice, and I drove the cars identically, so I'd consider my results useful. They don't reflect how the cars perform in all situations or when driving all sorts of ways, but it's alot more accurate than saying Al drove a HCH and got 46 mpg, and Bob drove a Prius and got 48 mpg, because Al and Bob can have completely different driving styles.

It's tempting to say comparing 300 HCH drivers to 300 Prius drivers minimizes such problems due to the sample size, but the problem is the distribution is not random -- the Prius might attract a completely different kind of driver than the HCH. For example, maybe "Bob" often drives with alot of luggage or a large family and went for the Prius for that reason. That can lower his mileage compared to a lightly loaded HCH. Or, perhaps Bob went with a Prius because it's a very conspicuous noticeable car, and Al went with the HCH for exactly the opposite reason. Al is then maybe less likely than Bob to be the kind of person who speeds alot, since that would make him stand out.

Of course I'm still just guessing on that last point.