I'd like to suggest reporting cost per mile to more decimal places.
In my personal spreadsheet, I have cost per mile set up to display to four decimal places, e.g. $0.0821. Having this figure reported to more decimal places allows me to recognize up or down trends much better than if it were simply reported to two decimal places (e.g. $0.08, as Fuelly does now on the Profile page). Consider that Fuelly's present rounding logic treats both $0.0750/mile and $0.0849/mile as the same figure...$0.08/mile. Yet there is a 13.2% difference, or $13.20 per $100 spent, between the former and the latter!
This would also help drivers better estimate annual fuel costs. A cost per mile figure given to four decimal places allows for an estimate of the fuel cost to drive 10,000 miles down to the nearest dollar (e.g. $0.0821/mile * 10,000 miles/year = $821/year). With Fuelly's current model, I can really only estimate 10,000-mile-cost to the nearest hundred dollars. If I drive say 30,000 miles a year, or if I am considering purchasing a new car and want to estimate fuel costs over multiple years, the difference between having a 2-place cost per mile and a 4-place cost per mile becomes even more amplified.
If Fuelly were to implement a 4-place cost per mile, I'd suggest reporting it as x.xx cents/mile, rather than $0.0xxx per mile (e.g. 8.21c/mile rather than $0.0821/mile).
Sorry, but we're going to stick with standard rounding and standard money units for our calculations. Having 3 or 4 decimal places next to a "total cost" figure looks odd when people are just trying to get a handle on their annual costs (where that much precision is unnecessary and distracting).
I understand that might lead to a penny here and there of difference but honestly we're not looking for scientific standards in our statistics tracking, the site is set up for the casual driver just wanting to get a basic handle on their fuel economy. I'll remind you that everything here is self-reported anyway and we'll never get absolute precision from all drivers so having some small rounding errors when extrapolating data is fine given the data wasn't collected in a clean room and is subject to all sorts of flaws.
Sounds like you're doing much more with your numbers and you're already in Excel with them, so you should continue using that and tweaking there.