As far as I understand, the voltage reference at the IN+ (1.2V) and the negative feedback force the voltage at IN- and consequently across the R1 to be the same 1.2V. R1 is then chosen to limit the current to 100mA. The right-hand side BJT acts as a switch and when a micro outputs a +5V, the transistor is biased into the linear mode and current flows through the LED. If a micro outputs a 0V, the BJT is in cutoff.
This is all very good and I know it works but I don't really see how this topology would be advantageous over a classic emitter follower with a single BJT that would simultaneously act as a current amplifier and a switch. Isn't this an overkill and unnecessary complication? The final device is going to use many multiples of this circuit and having unnecessary components would be very unreasonable.
Unfortunately, I have no chance to ask the designer whether he chose it for some specific reason, so I wonder whether I'm missing some perks of the circuit in question.
Also, I'm not sure how justified the use of C6 is at the emitter of U6.