I'm increasingly of the opinion that there's no such thing as a "no brainer" when it comes to giving instruction to a broad group of people (and "broad" might be as small as n=10).

You know the rules of program optimization? (#1 Don't #2 (for experts only) Don't yet). I suspect that illustrates the issue that experts are having here. Yes, in any program there are optimizations to be done that can change margins that matter. But without deeper experience even the average practitioner is unlikely to have good judgment about this. So the experts have to balance the effectiveness of higher-resolution explanations where details will get lost in translation (likely degrading any marginal gains to be had) with a low-resolution blanket rule that sacrifices potential gains for an acceptable average.

Several kinds of masks appear to be several degrees of marginally effective, but it's probably very difficult to give this guidance to the public without some segment of it hearing "I need N95 masks" and scrambling after supply, because "Any mask might be helpful to get me to be conscientious about touching my face and provide a marginal droplet barrier, N95s are for people spending time with the sick" is a higher resolution message that's subject to a lot of social/individual entropy.

Non-N95 and DIY masks are also very helpful, but the dogged "you only need to wear one if you're sick or caring for someone who is" advice discourages even this level of prevention, and makes the disease seem less serious and contagious than it really is. It's counterproductive to lie and say "masks don't work" when what you really mean is "we need N95 masks for frontline workers."