Is this for a one-off design, or for a commercial design? The reason why I am asking is that for a one off design, you can adjust the current limit to match the gain of the transistors. If it is for a commercial design, then you have to use worse case specs and with only a 20mA drive current, the transistors are near the thermal limits if you look at the full 2A through a single transistor.
So if you look at fig 5 of the data sheet for the DS2711, they set the current for 2A total. If you are charging one battery, then the 2A is going through one transistor. With a 20mA base current, the original FCX718 looks as if it will have a typical Vce of about 1V (2W power), and definitely less then 2V. If a transistor had a 1.4V Vce at 2 A, that would be 2.8Watts power which is way above the specs, even at 25C ambient. So I would assume that for a commercial design using the original FCX718, you would have to wind the current down to a lot less then 2A.
The FTZ951 has a worse case (minimum) gain of a 100 at 2A, 1V Vce. So the worse case power dissipation at 2A is about 2W. The transistor is spec'ed at 3W with a total of 4 square inches of copper for heat dissipation at 25 degrees, so we are getting closer to a reliable design. The FTZ953 is not specifically spec'ed at 2A, and with a higher voltage rating, it will probably have a higher Vce then the FTZ951, so I would not use it.
Richard