> predictable intervals
Yeah I'm going to guess it's methodological errors. Could be noise mining, could be something stupid like misapplied stats or a filter propagating results backwards in time. I suppose that meditator study rules out at least the intervals in their specific case but not really much else.