I find the exact opposite. I must parse the CSV myself, because every single one I get is slightly different. I've tried several libraries and none were flexible enough to cover all my use cases, though I haven't tried the one you suggested.
Currently, I have a GenericSeparatedValues trait that extends RegexParsers. I then usually specialize for the various "dialects" of CSV I need to support. However, the combinator parsers have a tendency to want to load everything up in ram and do a lot of subSequences and getting it to parse CSV that are hundreds of MB required some finagling.