Agreed. So I think the question is: Which effect dominates?
Both effects depend greatly on whether most other employers in the industry agree amongst themselves on whether to allow WFH: If everyone allows WFH, or everyone forbids it, there's no incentive to change employers, so these are stable equilibria, all other things being equal. Employers prefer the no-WFH equilibrium since (they believe) that leaves productivity highest.
Both effects depend greatly on whether most other employers in the industry agree amongst themselves on whether to allow WFH: If everyone allows WFH, or everyone forbids it, there's no incentive to change employers, so these are stable equilibria, all other things being equal. Employers prefer the no-WFH equilibrium since (they believe) that leaves productivity highest.