I don't think anyone knows. I don't have an answer.
Perhaps a degraded/dead cell or two that affect string voltages, versus packs that have no "dead pixels"?
I think this question ranks right up there with why I haven't seen (much) rated-range degradation after 47,000+ miles and nearly 2 years on my replacement pack, compared to the original rev A pack that saw degradation, charging to 90% @ 228ish miles. You can't say that I baby my pack - it gets range-charged frequently and will sometimes sit at 100% for a day (if I've forgotten to set it back to 90%), supercharged occasionally (but not regularly), always charged at 80A except while on-road, gets driven down to 5% or so occasionally, and my approach is very "don't care". The odds shouldn't be in my favor here - but the car still charges to roughly the same as when the pack was installed (linear regression estimates ~1.5 miles per year). Yet others with newer cars show ranges well under mine, and no amount of rebalancing and/or using the entire width of the pack restores any.
It points toward wk's theory that there's a bit of headroom the car can absorb and hides from the user, but we haven't established that with any firm and predictable evidence.