I'm quite confused. Trying to get a list of Canadian provinces and territories, i.e. the first-level provinces. So I'm looking for a subclass of Q10864048 (first-level administrative country subdivision), related to Q16 (Canada). The subclass I'm looking for is of course Q11828004 (province of Canada), which is indeed a subclass of Q10864048.
So, let's query the subclasses of Q10864048. This does not yield Q11828004.
Strange. Wikidata says that Q11828004 is a subclass of Q10864048, but query.wikidata.org says it is not. At least, query.wikidata.org finds 264 other direct subclasses of Q10864048.
Checking the SQID tool: https://sqid.toolforge.org/#/view?id=Q11828004. Yes, Q11828004 is a subclass of Q10864048.
Checking the superclass: https://sqid.toolforge.org/#/view?id=Q10864048. "Direct subclasses: none". So the SQID says that Q11828004 is a direct subclass of Q10864048, but Q10864048 has no subclasses?
And, how can a class have 14 total subclasses, when it does not have any direct subclass?
This does not make sense. Three tools, all saying something different, the SQID tool even contradicting itself.
One possibility would be that the tools are operating on different versions, i.e. a different data dump. However, the claim Q11828004 P279 Q10864048 has been in the data since April 2017, so it should be present in all data dumps.
What am I missing?
Note: It seems that query.wikidata.org knows about the relation Q11828004 P279 Q2879, and Q2879 P279 Q10864048. But this still does not explain the differences mentioned above.