Under the 2 version transistor circuit, the sense resistor will trigger, which turns on the first pulling the 2nd (that drives the LED) to gnd reducing current.
That works great for led's less than a watt, problem arises when applying to high powered LEDs, as the voltage rises the current will, the 2nd transistor and sense resistor is unable to pull the LED transistor to gnd while supply high enough current for a high powered LED ..
(I can show you that in a simulation it's easier lol)
So the PNP transistor (3rd transistor middle of the 3) why use it... the base of the pnp sits on the collector, higher the voltage the less saturated the middle (pnp) transistor is...
That means the base of transistor 1 is solely dependent on the sense resistor to limit the current to the 1st transistor (as normal in the 2 transistor combo)
So then lets turn up the voltage, 30v.
The sense resistor now shows over .7v and transistor 1 fires bringing the collector voltage down, because the emitter now has a higher voltage on the middle pnp transistor, it's now starting to saturate...
The pnp middle transistor then dumps its load out to it's collector, this in turn saturates transistor 1 bringing the base of the 3rd npn firmly down to gnd turning off the LED.
In a 2 version current limiter it can't reduce a high powered led's current but will fine for less than a couple of hundred milliamps...