QXmlStreamReader::readElementText(QXmlStreamReader::SkipChildElements) bug

I want parsing whole Wikipedia. I am interested with text element and I get it from tree:
+mediawiki
+page
+revision
+text
(plain text)
+ref
(plain text or more refs)

As you can se the text element can contain plain text mixed with elements like <ref></ref>. Problem is that QXmlStreamReader::readElementText(QXmlStreamReader::SkipChildElements) does return also ref! This can be observed on attached screenshot. You can see code responsible for parsing <text> element and debug output where plain text is mixed with <ref> elements. This bug is huge dissapointment!