I don't get it: if Google correctly identified that the melody was from a 100+ years old composition, and that it wasn't a reproduction of copyrighted content,

I don't get it: if Google correctly identified that the melody was from a 100+ years old composition, and that it wasn't a reproduction of copyrighted content, why did it raise a copyright infringement claim in the first place? It sounds like it has already a database of compositions it matches against, and the year of release is probably reported on that database. The solution may be as simple as not raising a copyright infringement claim on any match again compositions that are 70 years old or older.