On the uniqueness of maximum likelihood identification