I think the problem are the curves, not the zero segment (=‘stacked points’). When they are reduced to a straight diagonal line, they are not completely straight. They have a tiny inclination, which technically is a very different curve than the one it expands to in the other master. And it is different in such a way that compatible TT curves cannot be calculated.
Two solutions:
Retract the curve segments into the ‘zero point’ as well. Or:
Bend the curves a tiny little bit like in the second master, so that the curves bend into the same directions in both masters.