Right now, Distribute Over NPoints uses a random choice of evenly distributed points. So if there are an equal number of particles and points, it is extremely unlikely that all points will be occupied, and many will likely be occupied twice. Because it does not have the option to be seeded, this distribution changes each loop, so there are variations even when the initial location is seeded. It would be nice to have the option to use a random sample, so that if there are an equal number of particles and points, and the distribution threshold is 1.0, all points are occupied. No points would be occupied twice unless there are more points than particles. It’s a lot easier to control the location of particles through world offset if you can guarantee where they are.
I ran into this issue attempting to render point clouds in this thread.