There was a workaround for deriving the filename from the source filename
because characters like '#' didn't make it through.

Turns out this was just a bug some lines above where encodeURIComponent()
wasn't used as it should.

Use encodeURIComponent() when constructing the URL and remove the
workaround.