Disclaimer: I'm the original editor of N-Triples and implemented it in Raptor http://librdf.org/raptor/ both the N-Triples original and the 2013 version.
There are multiple answers to this but it's basically ambiguity. CSV can't distinguish between a URI that looks like http://foo.com/
and a string http://foo.com/
In CSV
http://foo.com/,http://foo.com/,http://foo.com/
this could be a triple
(URI http://foo.com/, URI http://foo.com/, URI http://foo.com/)
or
(URI http://foo.com/, URI http://foo.com/, Literal http://foo.com/)
N-Triples adds <>
and ""
for distinguishing these cases