Phishing has matured. The bad guys are now so adept at mimicking the actual emails sent by PayPal, that PayPal support apparently cannot tell the actual PayPal email apart from the Phishing emails.
PayPal mistakes own email for phishing attack [The Register]
PayPal admits to Phishing Users [eset.com]
I've wondered for years why the phishing emails were often so terribly lame. The ideal strategy would seem to be to read some actual emails from the intended target, and mimmic those as closely as possible. The traditional excuse offered by the security community is that the emails appear often to be generated by people who speak English as a second language, but that doesn't seem like it would be such a limiting factor, given the ease with which the translations could be corrected, even anonymously, using clever internet tricks, even fairly simple ones.
The real answer seemed to be that the text content of the email didn't much matter, as people don't read them very carefully. It appears to be from their bank. It's got a link. It says to fix your login. Click!
The competitive pressure, both from education efforts which make the population of victims more sensitive to potential identity theft, and from other Phishers seeking to exploit the same population of potential victims, seems to be forcing the emails to evolve to more closely resemble the target company's web site and actual emails. Witness the inevitable result: technical support can't tell the Phishing email from the actual company-generated email contact with their customer base.
Non-authenticated email is a zombie: un-dead, walking.