Good one Cary - quite an interesting question.
The original submission was based on the idea that the filters would be "standalone", eg. placed patient side of all the equipment in the Central Supply System. Thus they would "protect" the receiver, pumps, piping, etc. and of course the main goal - the community.
However, even at the meeting where this was discussed the question came up about the idea of one filter, one pump. My recollection is that there was general agreement that this would meet the requirement as well, albeit it would only "protect" the pump and community, so it was understood to be an inferior implementation.
You have correctly noted that there is no explicit wording to allow this (because there was none in the original submission) but equally nothing that prohibits it. The wording is adaptable and all the requirements in 5.1.3.7.4 can be met by the "filter per pump" variant (albeit (6) is probably fudged most of the time).
This underlying philosophy runs through the whole code. It is inherent in "two-source" (N+1) construction. If one pump is down for maintenance, we assume you will manage on one until it's finished. Downtime is always a planned event. As in this case, when you take down a pump/filter combination for filter service.
The annex and commentary in the Handbook of course follow the text. I agree it seems an oversight that at least the Handbook commentary makes no mention of the one filter per pump alternative implementation. I've not seen the 2021 Handbook yet, but sounds like it did not get any attention in the 2021 rewrite.
A good subject for the MGPHO Standards submissions forum.