A popular, oft-repeated quote appears in a biography. The note for the quote begins, "as quoted in...." You, the researcher, track down the part that comes in the ellipsis. The note for that quote begins, "as quoted in...." You track down the part in the ellipsis again. After following a breadcrumb trail of "as quoted in..." you finally find the document.
The document is an undated letter from "Anonymous" to your subject. At least the biographer lists a repository in the note (you can't always count on that). The repository no longer exists. The collection, however, does, so you track down the online finding aid to the collection in the new repository.
That repository has an index of recipients and senders that includes "anonymous" as one of the index items. If the document that was "quoted in..." is actually in the collection (you have your doubts, given some of the features of the biography with the document citation), it has been assigned a date or a sender and has not been cross-indexed under the recipient's name. It isn't there.
You want to use the quote for all of the reasons the "as quoted in..." biographers have used the quote, but you now suspect it to be a myth, the sort of thing that someone should have said but did not. You finally settle on avoiding the quote because that trail of "as quoted ins..." have convinced you that, even if not a myth, it is certainly a cliche'. You want to be original.
At least, you want to be as original as you can when using someone else's words to make a point.