I just reviewed all of the logs ever gathered for my Roadster, using my Tesla Graphical Log Parser. Found some interesting results.
In this graph, you'll see that my Calculated Storage Capacity (A-hr) starts out slightly low when the car is new, but levels out near 160 after I start driving the car (there was a bit of a delay before I got the car, long story). It stays there for nearly two years, then suddenly takes a big dip in February 2012. It then gradually recovers slightly. This corresponds to my full Standard Mode range going suddenly from about 300 km to 293 km. At the time I was rather confused about what caused that.
Well, here's the answer. On January 24, 2012 my Ranger installed firmware update 4.6.4 (it is now at 4.6.5). Nine days later I did a Range Mode charge but left immediately on completion and didn't let the pack balance. On February 12 I did a full Range Mode charge, allowed the pack to fully balance, then headed out on a road trip (to the Canadian International Auto Show). Along the way the car was drained to 10%, charged and driven several times, charged on 110V, left unplugged for two weeks at nearly full charge, and then the return trip with one charge stop. There is a three-day period starting at that Range Mode charge where the Calculated Amp Hour Capacity steadily dropped. At the time I wondered if draining the car so low had an effect, but the drop actually started right after that Range Mode charge. Nothing else about how the car was operated correlates with the drop.
So I have to conclude that the firmware revision was responsible. Once I did a Range Mode charge with full balancing, the car had a new baseline for the pack capacity, and then over a period of three days adjusted the CAC (and therefore the car's calculated range) to the actual battery condition. Apparently the range estimates prior to that were higher then they should have been, and this firmware revision updated the calculations.