MK, the example you mention wouldn't be affected by the mechanism I mentioned, because the OVERLOADED puzzle had Saturday's date and the UNCOVERED puzzle had Sunday's date. The control on puzzles with common words was designed when we had just two daily puzzles, and they started at the same time. So the comparison is only with a puzzle for the same date.
Now that we have three puzzles, with staggered starting times, there's more scope for similar puzzles to be active at the same time. Maybe the scope of the test should be widened to take in puzzles from the past couple of days.