I think this likely greatly understates the efficacy of a properly fitting respirator. If it's not leaking, it's likely to be very effective at capturing particles. And 95% is the minimum for a certain range of particle sizes, while large particles (the large droplets one would often be concerned with) will be even better. The key is to make sure all the air passes through the mask. Many of the quality masks out there fit very well (in the absence of significant facial hair) and ensure this happens.
Unfortunately, actually measuring efficacy of mask wearing is extremely difficult, and no one has done it (because there's no way to measure efficacy of mask wearing because no one wears them all the time). Of course it only makes sense to measure efficacy of N95 masks exclusively, since they are cheap, more comfortable than inferior masks, and there's really no reason to wear anything else. But I'm not aware of any studies of such things (actually measuring efficacy of mask wearing, excluding all periods of non-mask-wearing). No one cares about a study that includes periods of time when the masks were not used, and to my knowledge every study to date includes such periods.
The only evidence I have is that all the people I know who are serious about wearing masks still have not contracted COVID. It's not really evidence! I'll be the first if (when) I get COVID to come on here and say that I got it while wearing a mask, but I don't anticipate that will be the situation where I get it. It's much more likely to be in a situation where I am unmasked. If I get it through an unknown source, I will chalk it up to mask failure. I expect that I will know exactly where I got it, when I do get it.
View attachment 973833View attachment 973834
For the retrospective efficacy: This was for a test-negative case-control study, which I really don't understand even though I know how it works at a basic level, and I really have no idea how they remove the biases which might be involved. Just seems impossible to me.
Anyway. There are more slides there showing how the bivalent did, comparing VE measured via test-negative case-control, in various other scenarios, as well. And plenty more data on the new vaccine as well - but no efficacy measurements, of course.
www.cdc.gov