I have the following XML document:
@NOT_USED@
It doesn't look like the namespaces in this document are correctly declared - there should be xmlns:samlp and xmlns:saml attributes on the root node. In cases like this, Nokogiri essentially ignores the namespaces (as it can't map them to URIs or URNs), so your XPath works if you remove them, i.e.
doc.xpath(XPATH_QUERY)