It is a separate question to what extent those distributions are "physical"...
Simple attempts to establish that at least the size distributions obtained from a mixture of two samples with same composition and two very different size distributions, are close to the expected result, establishing some self-consistency of the methodology, if not that they are "physical" (I believe they are "physical" in case of size-only effect).
See for instance J. Appl. Cryst. (1978). 11, 50-55. This can be found also (in french) in a thesis : http://tel.ccsd.cnrs.fr/documents/archives0/00/00/70/41/index_fr.html (self citation...;-). Things have not changed a lot since these old times.
Armel
