Because it hadn't been established that he "took" them but only that he "may have taken" them.
Something being found in someone's home falls short of establishing that that homeowner "took" it, which would be jumping to a conclusion. If my stuffed bear is found in your home, is that in and of itself proof that you "took" it? No, it's at most circumstantial evidence that you "may have taken" it. So, if I publish that you "took" my stuffed bear without proof you really did, instead of publishing that you merely "may have taken" my stuff bear, which is all I have evidence of, then that exposes me to legal liability, to having to pay you potentially huge financial damages if and when you sue me for libel and defamation. Moreover, my using "may have taken" would not be at all suggestive that anyone else "took" my stuffed bear either because there are many, many ways my stuffed bear may have ended up in your home other than you "took" it or someone else "took" it.
By the way, this is an apolitical response. Do not mistake this as my defending Donald Trump. I'm simply explaining why a journalist and a periodical would properly report "may have taken" instead of improperly report "took" when reporting "took" with what that journalist and periodical had available to them at the time of publication would've been unfounded. Maybe "Trump took" later ends up getting proven, but until then, in order to protect their journalistic integrity and shield themselves from legal liability, it remains prudent for that journalist and that periodical to not report anything that they don't actually have proof of to back them up, and the absolute most they had proof of to back them up when publishing that was "Trump may have taken the documents," not "Trump took the documents."
