The title is not strictly necessary to track down a paper, so it can be considered superfluous information. Leaving it out saves space, paper, and money in the printed version of the journal. It also saves on shipping. (The names of individual journals are abbreviated for the same reason. Same goes for the page range being replaced by the opening page number.) That's the historical reason, as I understand it. Speaking of history, this practice goes back a long way, and was commonplace already in the late 20s and 30s, in journals published on both sides of the Atlantic*.
One might expect this to have changed somewhat with the rise of digital distribution, and particularly the newer online-only APS journals. It sort of has. In 2015 the print journals decided to allow titles in the reference list. As for the online-only ones, I haven't had reason to check the style guide all of them. However, I know that at least one of them, Physical Review X, even requires the inclusion of titles.
While Reviews of Modern Physics is still printed (and thus should prefer leaving out titles for references according to the above logic), it's a unique case in many ways. One of these ways is that it mostly publishes few, but quite long papers. Hence, the relative savings of leaving out the titles would tend to be much smaller than in e.g. Physical Review Letters.
Even with digital distribution, many of us still prefer reading a printout, so the paper savings are still appreciated. Another reason for preferring this kind of concise referencing is that it kind of fits how some (a lot?) of us think about papers (although I admit there might be a chicken-and-egg situation here). It's Brady's PRB from 2015, not whatever the title of that paper might have had. Also, titles are quite often similar to those of related papers, and can blend together, so they don't add much IMO.
*No special meaning is assigned to these decades, I just happened to have a few Zeitschrift für Physik and Physical Review papers from this time period on hand.