Where the implementations differ

Every row more than one implementation states, sorted by ratio, widest first. A difference is not an error. Competent teams rendering one method against different flow lists will differ, and most of these are modelling choices; the widest are worth a conversation, not a correction.

Published says what this list did about the row: `agreed` where they stated one number, `sole` where only one did, and nothing at all where the disagreement went to a queue instead.