At least for WP URIs containing (any of) the chars ":()", extracted triples end up with different subject URIs. Probably no (re)encoding should take place at all really.
For details, see
http://sourceforge.net/mailarchive/message.php?msg_name=4BE921C2.6080802%40dfki.de
The URI encoding is fairly consistent (http://dbpedia.org/UriEncoding).
However, there have been some encoding issues with URIs from external datasets (e.g. YAGO) which led to confusions regarding the scattering of resource data.