The problem with approval voting is that it does not give voters a mechanism by which to indicate partial approval. If there is a centrist candidate which many voters only slightly support, there would still be a media horse race over whether or not voters will interpret partial approval to mean 'do not approve' or 'approve' on election day, with a final result that may be unnecessarily surprising to participants. There are also examples on the wikipedia page of some organizations trying range-2 approval voting and abandoning it.

Range-3 seems like the minimum range desirable. It would allow voters to indicate whether they disapprove, were neutral toward, or approve of a candidate. The only historical adoption of range-3 I am aware of is the electoral council of the Republic of Venice, and I believed they used it as a constant component in their electoral process for selecting the Doge without abandoning it.

The costs of switching to something even better like range-5 may also be over-estimated. Range-5 seems like it would work with existing optical scan ballots. The cost of acquisition for range voting tabulatation code on the state level would also ideally be fairly low due to the simplicity of the preference aggregation method in range voting in comparison to IRV, and Maine is already undertaking a switch to the later.