I strongly suggest you abandon efforts to merge multiple trees where there are many people common in more than one tree.
Let me give an example. Let's say you have Jane Doe that the program identifies as a dupe. You merge Jane, taking the time to decide which fact (birth, death, etc) are the right one. In the meantime, one shows a husband: John Taylor Smith, b 1850 as a husband. The other file shows J. T. Smith, b 1860 as a husband. They are the same man. The program misses that they are dupes. The program imports the J. T. Smith as a second husband, where there was actually only one husband.
And, then, of course, the woman has a child Nancy Ann b 1852 with John Taylor Smith and Nannie b 1850 with J. T. Smith. They're different names for the same daughter. The program doesn't recognize the dupes and brings the two daughters into the two different fathers and mother. And the mess just multiplies and multiplies.
On and on this kind of thing goes and you end up with a mess.
If you set the parameters too loose, you will identify a whole bunch of false positives. If you set the parameters too tight, you will miss a bunch of dupes. In either case, you have to take the time to decide on a merge of a person which birth, death, etc facts should be the preferred one and which should be discarded or alternative - when you often won't have the info to know which should be preferred over the other.
You will be much better off to take the "best" file and visually look at the other ones and then make manual entries to enter preferable conflicting info or additional info. You will also be able to add a source for the original file (and its source) that this data comes from.
You can import "branches" cleanly, ie use File A, but delete branch 1 from it and import branch 1 from File B. Then attach the child from the imported branch 1 in File B to the parent in File A. But a massive merge of a two files with bunch of people in common, using the merge ability of genealogy software (built-in to FTM) is a disaster waiting to happen.
And this doesn't address the problems of merging two people together who look like the same person, but each showed different parents. Which is right? Or two different people? You will often not have the info to decide which is right when you being forced to make that decision in the software.