You must be careful at discarding some records that dont seem reasonable. For instance, I got an estimated mileage of 102 mpg on my bike (gsycaty, Sept-16, 2008). After looking at the data, I believe it is not really an error. It was a very small fuelling, and just after that, it gave 71 mpg. I tend to think that both are wrong, due to unequal incomplete fill-up. But the global aggregate is probably correct. Further, punctual errors tend to cancel, and on the long run, influence is minimal.
Of course, quality of data depends on quality of input, and reliability issues lie fully on the user.
What I would consider adding is the option to add a correction factor for the odometer. As you know, most odometers overestimate speed, (and so, also mileage), Some people do check this error with their GPS, and they know that their odometer is overestimating, say 10%, or 8%... Could this be added as an optio