On the importance of a correct choice of model for lanthanide ion binding in the interpretation of lanthanide-induced N.M.R. shifts: the adamantanone problem