At this point I'm 99.9% certain that we're looking at some kind of experimental error, but being unfamiliar with the details I'm not about to start speculating on particular things that might be wrong. Any possible error that I could think of would probably, if presented to the experimentalists, result in a rolling of eyes and a "Don't you think we already checked that?"

To put it in terms that people here can understand, suppose you're three days into tracking down some incredibly obscure bug and your mother asks "Did you try turning it off and turning it on again?"