Which one is that again? If it was me that suggested that, then I know which one...
It has to do with the (lack of) stability with a base-biased transistor (emitter current isn't temperature stable). It would be even better to use a different design that isn't so dependent on transistor parameters (I.e. Beta) (collector-feedback perhaps?)
I think what happened with the original is that higher case temperatures than we see in the DIY version further inflated the gain of the transistor (this is a bit of a stretch, admittedly) which made the circuit work more consistently.
I think the result would be more stability, but I guess I am the few people that sees this part of the circuit as not being robust - likely because I seem to have the most problems.