He's likely right, it's a best practice that may be wrong in some cases. Though generally, you're better off going with a best practice and then A/B testing alternatives.
The reality is that you're limited in the number of A/B tests you can do, both by traffic but also by your own time and resources. So you can run a test against something that is generally accepted (like what text works on a button) or against more important changes like headline copy or page layout.