That's very odd, I guess I can't rely on elevation tracking either...This and this agree, and this is 10% off (100 feet difference over 1000 feet). I think I'll stick with mapmyride since I've already used it a bit. I just mapped the same route with bikeroutetoaster and it agrees with the first two, so me thinks those three are the most accurate.
It's not a simple matter to know what the "right" number is. This Excel table http://www.caltriplecrown.com/ClimbingDoubles.xls of altitude gain readings from California double centuries over many years from many cyclists using a wide variety of devices shows just how much variability there is in raw readings from barometric altimeters, let alone how GPS mapping algorithms decide how big/small an increase must be to qualify as a sufficient change to incorporate into the total gain.