All should work but both transistors should be PNP so it's basically a darlington pair.
The concept of the 'wrap around' transistor is that they drop the bias voltage for the pass transistor across the resistor in series with the regulator input pin. At first, it seems counter-intuitive to monitor the input of the regulator rather than the fixed output voltage but it does work, and it works well. The values have to be calculated so the input resistor still allows some current to flow through the regulator, I usually design for about 75% of its rated current to flow through the regulator with Vbe (say 0.7V) across the resistor. That means that up to that 75% figure, the resistor drops less than 0.7V and the transistor doesn't conduct. Now, if the load current increases, the transistor starts to conduct and everything above that 75% figure passes through the transistor instead. The voltage is still fully regulated because the input current to the regulator depends upon it's output voltage.
The big advantage, and reason for running the regulator at relatively high current, is that if it goes into thermal shutdown, it stops drawing current and the pass transistor shuts off too. A good thermal bond between them also helps.
The downside of that configuration is how it reacts to a sudden short circuit, before the regulator has time to shut down, there is a potential for a high current to flow from the reservoir capacitor through the small transistor and straight through the main transistor base junction.
The schematics would work better if the small transistor was removed completely and a darlington PNP power transistor used instead. If you try that, remember the voltage dropped across the input resistor has to bias TWO PN junctions so calculate for 1.4V drop at 75% regulator load.
Brian.