I reckon it depends on the level of information. Given perfect information (and rational voters), it’s impossible that there is a Condorcet winner other than the approval winner. So how do we decide the best candidate in the absence of a Condorcet winner? Certainly not the runoff, as all that tells you is which candidate has the direct beat-path and which has the indirect beat-path. Maybe the approval election we just held (i.e. the runoff is pointless)?
Granted, the likelihood that a second-most-approved runoff winner is a Condorcet winner increases as the level of information diminishes, but the burden of proof is on the proposers of the more complex system to demonstrate that that likelihood is sufficiently high to justify defeating the approval winner.