the diamond princess is especially poor evidence, as it was an after-the-fact investigation (with heavy motivation to find something and gain early notoriety), not a study per se. what likely happened was that there were multiple social infection vectors (not of the long-distance air variety but more the drunk chatting variety) and the crew and passengers distorted their testimony to limit their own liability exposure. the diamond princess case was also the source of the "virus lives for weeks outside the body" mistruth because pcr tests readily amplify partial, dead viral dna/rna just as well as live virus (it has to be 'killed' to be amplified, after all).

but yes, viral concentration is higher near an infected person, but it falls off rapidly (to something like the 3rd power iirc) with distance and time, especially the latter of which being why incidental exposure in public places is unlikely to be material. it's also something a small minority of people have at a given time (on average less than 1% of those around you). it's mainly socializing for extended periods of time where masks can potentially help reduce infection (that's when we're also most likely to let our guard down).

i'd go so far as to posit that sociability (via variables like social graph size/breadth) is likely more correlated to infection than (the lack of) mask usage.