The analysis found a 98 percent correlation between the two datasets.
"That's 100 percent confidence level, it's our data," DeHart said.
diff ours theirs | xargs fbi_arrest_warrent_generator.py
edit: [#] that seems a bit aggressive, but I am not aiming at the parent post here, apologies if it reads badly. I think I mean these guys would make it onto Americas Dumbest Hackers TV special if that were the case.
It seems to me that the easier way to achieve this is either by randomly modifying the order of lines, which they didn't do, or to add/remove some bytes.
It could be the 2% diff between the two files : bruteforcing the md5 hashing by removing some random lines till they got a md5 digest containing 1337.