Thanks all. I think it's pretty obvious that FTDNA is the one with the issue. I will put the above in my other post for the benefit of FTDNA programmers. The fact of the matter is that for FTDNA's purposes they don't really need the alternates. They're just displaying a very basic ancestry tree. Plus, they don't have any way to display them. Any research would be done elsewhere and then a new GEDCOM uploaded. So IMO all they need to do is modify their import to take the preferred (first in file) and ignore all the subsequent facts of the same type for the same individual. That's what I'm proposing they do.