On the validity of the boson method for transitional nuclei