A horrible little ElementTree gotcha
What does this print:
from lxml import etree
doc = etree.fromstring('<a><b><c/></b></a>')
newdoc = etree.ElementTree(doc.find('b'))
The answer is: [<Element a at 817548c>]. The first point to note is that xpath() against an element is only relative to that element: any absolute XPaths enumerate from the top of the containing tree. The second point is that the shallow copying of etree means that _Element::xpath, unlike _ElementTree::xpath, evaluates absolute paths from the top of the original underlying tree! So even though there's no <a> in newdoc, an absolute XPath on a child element can still reach it.