I would definitely NOT just delete or even merge any duplicates shown by FTM with a high score, even 1000, without checking them very carefully first.
Apart from the danger of losing citations when deleting duplicates, even merging isn't always appropriate because FTM's duplicate test is far from reliable in my experience. At the moment it's showing me four duplicate pairs with a score of 1000 in my tree, but not one of them is really a duplicate - they're eight quite distinct individuals (see screenshot). A common instance especially in the Victorian era with high infant mortality is where a child dies in infancy, and the parents give the same name to another child born a few years later, sometimes even two or three. Another example is where you know of the existence of some children and have entered them with estimated birth years, but you do not know their first names. FTM will often incorrectly show these as duplicates also, with a score of 1000.
I do occasionally use the FTM duplicate check, but purely as a cross check for review.