Earlier quoted context omitted.
tl;dr - There wasn't really a conundrum to explain related to mask-effectiveness at different levels. Rather, it's more like a lack of good data, as prior studies being cited were ridiculously weak. --- This seems like a largely helpful characterization. However: > So they are left with a conundrum that places which use surgical masks seem to be better off, whereas randomized control studies of surgical masks show li…
The problem is people are not lab mice, you can't purposefully expose a test group to a virus to get clean data. So you either have to work with noisy data (as the previous studies you criticize) or turn to modelling (as this study does). This doesn't negate their value in either case. No need to be so picky.
The prior studies were neither well-done nor directly relevant. As such, they didn't constitute good evidence on the topic of mask-wearing for the recent pandemic. The proper understanding would then be that the effectiveness of masks was little informed by such studies.