Should a numeral before the comma count as a name word? (Dumas VI, Jr. gives no family name, Dumas III, Jr. gives family Dumas) #893

Background. To read a part before a comma, the parser counts its name words (rules/C3). AGENTS.md describes the count this way: a particle run and the word it attaches to count as one word; a connective join and a bound given-name pair count as their words (#613, so Ortega y Gasset, Jr. reads as a whole name); and a suffix beside the surname counts as no name word. A shape-only numeral (VI to X) and a chained salutation's connectives are counted as name words today, though they aren't:

The difference shows only when the part after the comma holds no given name: a generation word, or since #890 a gated agnomen (Dumas VI, Filho). #890's review found it, and it is older than #890.

How often: among comma-bearing local-corpus names, a two-word head ending in a numeral is common in FEC (2,431, nearly all SMITH III, JOHN), but VI to X there is 5 in FEC and 1 in ClinicalTrials.gov, and none of those has a part after the comma without a given name (counted 2026-10-25).

Questions

1. Should C3's count ask the numeral predicate (reads_as_generation, Ask one predicate for every roman-numeral rule (#752, phase 2f) #886) and count a numeral it reads as a generation as no name word, as it already counts a listed suffix? Then Dumas VI, Jr. reads as Dumas III, Jr. does.
2. Should a chained salutation's connectives (Mr. and Mrs. and Dr.) count as no name words? Or does that conflict with Rethink the suffix-comma rule (C1): decide the comma's structure once, after the words are tagged #613's decision that a connective join counts as its words?
3. Or leave both, as shapes nobody writes? Record that in decisions/C3, so a later review doesn't rediscover them.