Load prefixfile from same dir as infile. Use more recent Path::Tiny (needed for resolving samedir).
I disagree: I find it sensible that input files are tied together but output file is tied to current working directory.
Typical (to me) use case: Source data is in a read-only location, so I process it from another readable location, expecting output to be spewed to where I stand.
Guess I can live with that.
Please see https://github.com/VladimirAlexiev/rdf2rml#near-term "batch" for another promising output "model" (I also cited it in another comment somewhere here)