Actually, to be really strictly accurate, the situation with is a bit more complicated. Its valid html and should be valid xhtml. If you are serving xhtml as html then there should be no problem. If you are serving xhtml as xml then the xml parser may well only allow the 5 named entities which are valid xml. The reason for the problem is that an xml parser trying to parse xhtml would have to query the xhtml doctype to ascertain what entites were defined there. If it doesn't do that then it will only regard the allowed xml entities as valid. Since the approach is different from browser to browser it is safer to treat xhtml as if it were xml and only use the 5 named entities permitted by xml which are <, >, &, " and &apos. It should however be noted that ' it not valid html.
MattF is probably partly right about the validator. They recently updated it so it could check documents served as xml and it probably needs some fine tuning.
The short answer is if you treat xhtml as if it were xml and use numeric references then you will never have a problem. The alternative is to use utf-8 and enter the real character.