What I do about spurious relative connections is check to see if they're common errors. Do they show up on lots of public genealogies? If so, I put a stern note *at the top of each profile* warning that this is a common error, and please don't attach her to them.
Here's some good examples of profiles where that's been done. I prefer to say "erroneous" rather than "disputed" when the connection is clearly an ERROR rather than an ongoing dispute.
Susan French-3613 -- died in Britain at age 24, therefore NOT the immigrant wife of John Bridge.
John Bridge-134 -- had a LOT of spurious connections, including the above Susan. This is an excellent example of documenting such connections. (Here they're called "disputed" -- I prefer the franker but harsher "erroneous" for completely unsourced or disproven relatives, but I believe I'm in a minority, so take your pick.)
I notice Susan French-3613 isn't back linked from John Bridge-134, and Elizabeth French-6400 who PROBABLY originated with Susan, doesn't have a comment documenting her likely fictive nature or non-relationship to John -- she also ought to be in the {{Uncertain Existence}} category.