Nope, this is a feature class, I have replicated the same process. But, I still see few gaps in the output which shouldn't be there, like why is it picking few points from one cluster and few from another, while ignoring the other available (nearest).