A better efficiency is fully accommodated over time by the rising difficulty, so we are effectively time-invariant in the long run (efficiency vs difficulty), i.e. the halving interval is irrelevant if we still hit the efficiency wall before the halving. In other words, your assumption would hold true only if (and for as long as) the efficiency outpaces the difficulty. Previously, this condition had been satisfied by the new more efficient equipment hitting the market on a pretty regular basis...

But now we seem to be well past that point