We should find some interns with spare time to do actual count of particles in a sample field.
Several years ago David Francis of Metal Improvement Company made model estimates that seemed to support the number of particles for each media size maintaining a (fairly) constant percentage.
I'll search for that document (in my spare time)
