I did look at the graph you posted and I believe those 10% are the CV (constant voltage) part of the charging proses, but I didn't understand the text well enough to comment on it.
The reason why the pack doesn't balance when not charging (the ones in series that is) has to do with the way the balancing is done. When a cell reaches its final voltage of 3.65v and the rest is lower the
BMS solves this by discharging the highest cell over a resistor down to a level where is can safely be charged again, then it continues charging the whole pack. This is a proses that is looping until all the cells are fully charged. If you would try to balance without a power supply you could only do the first part and discharge the high cells and all you would end up with is to discharge the whole pack down to the lowest cell. If you needed to ride somewhere that day you'd be out of luck.
Some more expensive charger can charge individual cells instead of draining the highest cell and charging the whole pack. But to do this you need a regulating circuit that can regulate the voltage down to 3.65v before feeding it to the low cell. Absolutely possible, but more expensive and since balancing is only needed occasionally (about every 10th charge) I for one wouldn't pay the premium.
The alternative method I guess you where thinking about would be to steal the current from the highest cell and give to the lowest cell. You could build a
BMS that does this, but you would have to disconnect those cells from serial connection and then parallel connect them (if you try to parallel connect them while in series you'll end up with a short circuit). To do this you'd need transistors for every cell capable of delivering the full discharge capability of the pack (30A for a MPIII) when using the pack, and not only would this be a very expensive
BMS, you would also have a problem with heat from the internal resistance of the transistors. If you connect a small resistor in parallel with the highest cell instead you only need a transistor that can take the discharge current over that resistor as it will be in the off position when using the battery (plus it is a closed circuit so the transistors would only be subjected to the current over the balancing resistor even if it was in the on position when using the pack)
As for the question about the higher capacity cell in a pack. As long as this cell is only in serial with lower capacity cells and not it parallel I believe it would be charged to its full capacity given enough time. If you have a pack where all cells are 5ah except for one cell that is 10ah and they all where charged to 5ah capacity you would, in a perfect balancer, need additional 5ah delivered in CV mode. This is because all the 5ah cells would need to be discharged 5ah while the whole pack is charged again. As I showed in my last post the charger is very slow in CV mode and I would expect it to take about 10 days to balance it not including the time it takes to discharge. For a
BMS that can only discharge one cell at the time it would probably take several times longer. A smarter
BMS would discharge the cells so low that you could use CC (constant current) to bring it up again, but no battery packs are build like this, so no
BMS is going to be designed to deal with it effectively. After the initial balancing though you would probably never get that high capacity cell too much out of balance with the others (that is it will still have 5ah capacity left when the LVC kicks in and therefore only needing the same 5ah of charging as the rest of the pack)
If your pack had cells that where 4ah out of balance in your pack it would have taken a very long time to get it balanced with the
BMS.
This post got a little long and I'm not en expert in this field, so I expect there to be some errors.
