Let's take a hypothetical (worst) case where you have only two riders. Rider A recorded 1 tank at 30 MPG, and rider B recorded 30 tanks at 50 MPG. Ignoring the rider your graph would show an average MPG of 40 MPG, and your viewer would only know that their fuel economy was as likely to be 30-50mpg, and anywhere in between. The problem with your site's data seems to be that you have two main categories of reporters: those that report 1-3 tanks, and those that report them all (yeah!). The problem here is that the underachievers with few tanks outweigh the overachievers which have more data.
Let's use the data from one of your bikes: Tiger 800XC.
That model has 17 registered bikes, but of those only 13 have recorded one or more tanks. So your distribution plot shows those 13 tanks (see
http://cmumford.com/vstrom/BikeDistribution.png). Looking at the graph I think that I'm most likely to get 40 MPG.
But in reality you have 118 tank fillups and in actuality the most common fuel economy is closer to 42 mpg (see
http://cmumford.com/vstrom/TankDistribution.png). I think that plotting per bike or better yet per mile will yield a smoother distribution and better results for the user.