I think it's worth highlighting that there are two separate (but related) issues:
- What allows the algorithm to be most accurate
- What's best for the battery pack longevity
First- The algorithm:
It must estimate overall capacity of the pack based on some direct measurements (overall min/max voltages, module min/max voltages, charge energy delivered, temp, etc...), as well as some things that can't be measured at rest state, but must be observed as the pack depletes: amount of current delivered, voltage drop over time, etc...
Some of those items (specifically voltage limits) can only be truly measured if the pack is as close to possible to 0% and 100% respectively. It is also at the upper limit of pack charge that it has a chance to most effectively balance the modules, which also eliminates any capacity loss due to a module not having a chance to balance up.
Thus, for the algorithm, full range charges and driving till shutdown, however impractical, would allow for the greatest accuracy and least range loss due to imbalance.
Secondly- The pack:
The health of the pack is best with fewer "full-equivalent" charge cycles, fewer max charges, less excursion in to low charge states, and cool temps. In other words, operate it "shallowly" in the middle of it's range as much as possible.
Although this can lead to pack imbalance, that's not harmful to the battery, just inconvenient to us. The battery management system uses the lower limits of the lowest brick, and the upper limits of the highest, to determine what to set the upper and lower discharge and charge limits, so the pack is always protected.
As can be seen, the two operational schemes are at odds with one another: Pushing the pack to it's high and low limits often is best for the algorithm accuracy and slightly improves capacity due to balancing. It's also the hardest on the pack health. This is a bit frustrating for owners (and likely gives Tesla headaches in trying to tune the display behavior).
For me: the pack health is a priority. I'm not gonna sweat a few miles of variance. My daily driving requires only a fraction of my capacity, so I normally charge to 60-70% and drive it down to about 30-35% on weekdays. On weekends I charge to 90% and may drive it down only to 70-80% at some times, or all the way down in to the 20's at others. For the occasional road trip I'll range charge, but that's becoming rarer with more superchargers sprouting up.
I likely have some inaccuracy in my range displays. I may also have some imbalance. I'm OK with that if it means my (expensive) pack is happier. My 60% charge is right at 147 rated miles. I can often drive a couple of miles before losing any estimated range, so I'm pretty close to 25 miles per 10% of range charge. (my 90% charges are at about 221, so again close). That would put me at just shy of 250 miles for a range charge (which I haven't done in months).
I'd guess that if I did a series of range charges and deep discharges, I'd gain several miles back due to balancing.. and also some due to accuracy as the algorithm learned from the greater sampling extremes... but I don't need it and if my trips are that "close to the edge", I'd probably plan alternate charging stops anyway. The number on the dash isn't enough to get worked up over...