The BMS monitors the voltage of each P-group (21 per module, 84 total). Each P-group has 6 18650 cells. However, the system looks at each p-group as being ONE unique cell, not 6. If during discharge ANY p-group reaches the minimum voltage (the ave voltage of the 6 18650 cells wired in parallel, or the 5 or 4 or . . . depending on wirebond or other issues) as set by the firmware the BMS will shut down the pack (open the contactors) to protect that particular P-group from becoming over discharged. Li-Ion cells can be permanently damaged when over charged or discharged, that's why there's parameters for both.
In this case it looks like the bad P-group simply lost voltage faster and/or charged slower than the others and over successive ride/charge cycles is now at an open voltage much lower than the rest. Since it's an issue that will trigger a low voltage cut-off the effective capacity of the pack will be severely reduced as there's no way the passive balancing system can (or want to) dump that much energy from the good cells. Even with an active system I doubt the system could keep the bad P-group "pumped up" enough to not affect the effective capacity via low voltage cut-off.
The system also keeps track of all the energy going into (charging) and out of (discharging) the system, ie coulomb counting. By using this data and applying various algorithms the BMS can actually predict how much range (effective capacity, not necessary capacity as indicated by each cell) is available at any time. This helps the rider gauge his/her decisions on the trail/track and from a long term standpoint can assess the overall health of the pack. For example, as the pack ages it loses capacity. End of life is typically 80% (20% loss) of capacity. Without long-term re-calibrations you'd be running out of charge and the display would still say you have 20% pack remaining. You wouldn't want that. By the same token if the system senses a low P-group voltage it can use it along with all the other data collected to calculate about how much effective capacity you can expect and it'll report that via the display. That's what you want as it wont leave you stranded out in the middle of nowhere with a dead bike and a display showing 80% pack remaining!
Having said this it would have to have been an issue that's been getting progressively worse over time. It's possible Alta revised some of the algorithms to better indicate effective range via the display and you might see a sudden shift upon the firmware update if your bike had some p-group issues, I'm not that familiar with the software end of things.
