Well, it sounds like statistical significance counts in the one case (when it supports evolution), but not in the other case (when it demolishes evolution). You are certainly correct that it has been argued there is a strong, statistically significant phylogenetic signal. But those methods entailed silly (strawmen) models to compare against. What Ewert shows is a more interesting comparison of models.
I think what is happening is you are resistant to contemplating a new model, even though it is evaluated using standard methods and practices.