A critique of the term value approach to determining molecular constants from the spectra of diatomic molecules