In his book, 'Ergativity', R. Dixon makes a prediction on page 95:

if there is a split between bound and free forms, the former will follow an accusative pattern, the latter - an ergative pattern.

The basis for the prediction is the Nominal Hierarchy wherein personal pronouns occupy the left-hand side of it which reflects the fact that they are more likely to follow accusativity when there is a split.

I have not found any data pertaining to the matter in question on WALS. So my question is:

Is there a language that has the kind of split opposite to what Dixon's prediction leads us to expect?

Thank you!