They may well be executions of the same basic design, and even performed in the same factory, but there are always subtle differences between sets of tooling, (both for the plastic base moulding and rail forming) even if made by the same people or team. Very small differences in polymer quality, injection moulding set up and assembly process can introduce further variation during production too. Trust me on this one, years of industrial mass production quality assurance experience speaking.
For larger models in particular, with the finer wheel standards that are now normal, the simple fact is that set track pointwork is pretty much on the limit of what such models can negotiate reliably. Just as Sol posted, it only needs some small factor varying in gauging or flatness of the point, not visible to the unassisted eye; and of two identical looking points, one will perform reliably, the other will not. You can try measuring for differences, and experiment with corrections, for example one common suggestion is to put a thin shim on the inside faces of check rails to ensure the wheelsets are constrained better to follow the right road through the crossing. But that is only possible if all your stock has similar standard finer wheels. (That's the bind the manufacturers are in, their set track is expected to work with a large legacy of older product in owners hands, much of it on coarse wheels; and it also has to work with the more recent products.)
The real answer (if possible for you) is to move to a track system offering easier radii and finer clearances on points. The UK RTR stock manufacturers don't actually sell such products, so they don't advertise this ...
