use Poisson disk sampling in the Meshlab to sample the points from the meshes

The main parameter of the Poisson Disk sampling is the radius of the disk, and guessing the exact radius that implies a precise number of samples is quite tricky.

By checking the parameter exact number of samples the filter will do a bisection search on possible radii to get a more precise approximation of the required number of samples.

If you are interested in just good sampling over a surface (and not to the empty disk property of a Poisson Disk sampling) you could also use the Voronoi Sampling filter, that, while it is slower, it generates even point sampling over a mesh surface: random sampling of the points over the surface followed by a Lloyd relaxation that distribute the points over in a very even way.

On the other hand if you just need a few random points over a surface you can just use montecarlo sampling, but you do not have any guarantees on the sampling (e.g. you can get gaps and concentrations).

Here is a sample colab that with a small python program showing the above filters in action: https://colab.research.google.com/drive/1sHz28IHw3VfO2s6uim4CwaSc05idraP_?usp=sharing