I usually adjust my dedup settings to look at other fields – pages, vol, author – dedup those, then maybe titles and not authors, as sometimes some databases use diacritics, for example. It means rerunning a couple of times, to catch the variants, but prevents too much manual review. Sometimes I even do just vol/pages, and that also brings up records not really dups, especially those that were uploaded in their prepublication database entry. I then select the journal articles and ask Endnote to update them, so they get the vol/pages info.