Setting aside the arguments of sources with Phoenix, I'd like to prod your brain on this:
Very early in the pandemic, before things were politicized, I recall discussions of mask wearing, where the general message was that professional masks designed to stop fine particles could prevent transmission of a virus through them, while simple cloth masks cant do that, the particles are too small. However, the virus being suspended in moisture, which cloth tends to absorb, gives at least some capacity for cloth masks to slow the spread by catching droplets only if they are cleaned and only for short periods of time. The recommendation therefore being to wear a mask briefly if moving through a crowd or interacting with a vulnerable population, and otherwise either replace it frequently or don't even bother.
It seems to me that understanding of masks cleanly explains the seemingly contradictory results of different tests: immediate physical tests of masks tend to show them working well, even cloth masks. Which makes sense, if they are clean and dry when the test begins. But on a large scale, comparing populations by mask mandates gives little confidence to the idea that they significantly slowed transmission, which makes sense if everyone is just wearing the same dirty masks for long periods of time.
Does my memory and interpretation of events seem plausible to you?
Basically you're on the right lines.
The virus floats around in particles in the air. In an enclosed space with infected people, there will be virus present, and the more infected people and the more time people spend in that space, the worse it will be. My way of viewing it is that someone's chance of picking up the virus would probably be similar to a drug dose-response curve, where dose is a combination of virus concentration and length of time of exposure.
The obvious aim of masks is reducing the virus breathed into the air and preventing entry of virus from the air to the lungs (the former believed more important), thus reducing the dose. However, masks are not perfect and someone in a busy, enclosed space for a long time is very likely to become infected irrespective of mask wearing. For this reason, yes, the parameters of the study and activities of its participants may be very important to the result. Test indoor workers in an enclosed place working 7-8 hours a day in and out, masks might turn out to be low effectiveness, even the best ones - at best delaying the inevitable (although that will still slow spread). Test people going to a supermarket for 20-30 mins a day, they might be a lot more so. If almost everyone is sheltering at home and only meeting outdoors, masks may appear useless because there's low transmission. (Although, of course, if one person in a household gets it, the high length of time and enclosed space then applies and everyone in the household gets it, masks or not.)
There's also another factor to consider, which is the amount of virus someone is exposed to. For many illnesses, the higher the viral load, the more severe the symptoms and the more infectious a person is to thers. This is not totally nailed down for covid, but there is plenty of evidence in support. Thus masks may also reduce the healthcare burden that way, too, by reducing the viral load at initial infection.
Other key elements here then include population level behaviour of adherence: the proportion of people who mask up can have a huge effect. 20% masking may be borderline useless, where 70-80% masking may be highly effective. One of the other major problems is that a lot of these studies don't actually know what the adherence is, and in some cases, they are likely not to be good. Imagine a drug trial where a researcher is trying to see the effectiveness of a drug, but a high proportion of the patients aren't actually taking the drug and the researcher doesn't know precisely how many or who. For many drugs, it is doubtful a difference could be seen from placebo even if the drug works. Thus a study can simply be incapable of observing whether it works or not (or at least, not without more participants than is practical to accomplish).
So mask wearing may indeed be of low use in some circumstances, and very useful in others. But if too few people use them, it all but guarantees that they cannot be observed to work.
My issue with the sources Phoenixmgs is relying on is that some of them (e.g. Marty Makary) are just a load of bunk. Mostly, however, they represent a very specific sector of the research community of people heavily into "Evidence-based medicine" (EBM). EBM is based in a series of reasonable beliefs, however, like any set of beliefs, it collapses into absurdity when taken to extremes. And taking to extremes is what these guys do. I am particularly concerned that they have become far too emotionally or ego invested in this, to the point they have ended up doing things like acting as de facto anti-vaxxers. One factor to consider is that being very active in forms of social media, I think they cannot help but be affected consciously or subconsciously by their viewership, including through income such as paid subscriptions on Substack. They are not pulling viewers who are neutral, and it becomes very likely they are tending to push the messages they perceive their audience want to hear.