Is there a proton-neutron interacting boson model rule for M1 properties?

We investigate the robustness of the purported correlation between the g-factor ratios in the y and ground bands and the corresponding E2/M1 mixing ratios in the proton-neutron interacting boson model. We show that this correlation is dependent on the choice of the Hamiltonian and can be transgressed, when the parameters are chosen appropriately. The recent M1 data in ¹⁶⁸Er, which do not exhibit such a correlation, are analyzed in the light of these results.