Reply to comment:

Stéphane Bortzmeyer
I don't have time right now for analyzing the addresses (these are public measurements so someone else may do it). But I don't think it would be a good idea to exclude probes based on the type of address. Global and "normal" IPv6 addresse may have connectivity problems, too. Exclusion from measurements should be based on actual tests (probing the Anchors), not on guesses.