Yes, that design is ludicrously bad. There is no compensation for the varying output transistor Vbe drop with load current, nor for the emitter resistor drop, using Ib as a proxy for Ie for the current limit is a dumb idea in the first place due to the large variation of hFE with Ic, and the current limit circuit doesn't - its just an overloaded Vbe multiplier.
Boosting its output with an emitter follower is just plain dumb - see Figure 23. High-Current Adjustable Regulator Circuit in the Texas Instruments LM317 datasheet for the right way to go about it.
However, adding variable current limiting to a LM317 that works right down to 0V out isn't easy, so its probably best to throw it out and start from a regulator that's easier to apply limiting to, or has it built in or drop back to the classic two opamp CC/CV control loop solution.