The assumptions are dependent on the gene/proteins that are observed. If the protein is not preserved across species then Lynch’s model is the most useful. If they are preserved then Behe’s model is the most useful.
For beta lactamase in bacteria I would use Lynch’s model for beta catenin or WNT 11 in vertebrates I would use Behe’s model.