Well that is how constant current circuits work! But when you short it, you are providing a very low resistance path. Say when you shorted it it was 0.1 ohm. At 100mA, V=IR, so the voltage drop across this resistance is: 0.1 * 0.1 = 0.01V. If your power supply was 15V, then the voltage drop across the regulator would be 15-0.01 = ~15V. There is a dropout voltage across the regulator though, which determines the maximum voltage output. When you shorted it, its Vout was pretty much 0V. But if you put say a 470ohm resistor across it, it will try to force 100mA through it. Again V=IR, = 0.1 x 470 = 47V. Now you're powerign it with 15 so it can't go higher than that, but if you do this and measure Vout with respect to ground, I suspect it will be around 14V.
Also, when will you short the output? Sure the total voltage across your batteries will be 12.6V. And if charging at 2A, the load will be 12.6x2 = 25.2 watts. If you were to turn on all three shunt resistors/transistors, then they will dissipate this 25.2 watts... but balancing circuits aren't really designed to permanently shunt a battery - I believe they are meant to periodically reduce the current flowing through a cell, so it charges slower than other cells that aren't shunted. This allows those slower cells to 'catch up'.
I looks like many circuits don't actually 'balance'. They simply charge all three cells with constant current, then shunt regulate the voltage across any cell that has reached its target voltage (4.2V). So really, it just 'turns off' charging of a cell once it has been charged. I believe your hobbyking charger only had two shunt resistors because it works in a similar way. That is, as each cell reaches 4.2V, the resistor is switched across it. But once two cells have been charged, it only needs to wait for the last cell to charge.. once this reaches 4.2V, no need to shunt it with a resistor, because all three cells are done! So for that charger you need cells-1 number of shunting circuits.
The downside of this is power dissipation. I suspect the hobbyking charger doesn't use a very high charging current, probably <1A. The resistors aren't big enough to dissipate more than 3-4 watts. For reasonable charging times (7 hours, over night) you will indeed need 2A+, doubling the dissipation, making resistors significantly larger, and requiring larger heatsinks on transistors
I would use a shunt circuit for each cell, with a higher value resistor, so it steals less current from the charging battery. And your micro should periodically check the voltages of each cell, and turn on the shunt periodically for the cell with the highest voltage.
It's mean to 'balance' things out, constantly adjusting the charging currents to keep them all, more or less the same.
I got the impression you hadn't thought too much about power dissipation. When dealing with currents of >1A, every volt across a device, will dissipate over a watt. A TO-220 package, without a heatsink, can dissipate, at most, around 2W. For resistors, if you have a 2.2ohm resistor across a cell, that is used to shunt charging current, the current is 2A, voltage is 4.2 = 8.4watts. A 10W resistor would get extremely hot if this was switched on for any length of time.
I have never deisgned a balancing charger, and frankly don't know much about them, I'm just guessing here, and saying what *I* would do in that situation.