Isotope effects on isotope effects. Failure of the rule of the geometric mean as evidence for tunneling