Those are extremely close, 0.7% difference, so from the optimizer's perspective, they are more or less equivalent. Within 1-2% is our goal with the optimizer.
The optimizer is able to rank a LOT of gear combinations very quickly (up to several quadrillion in just a few seconds). That speed costs you a very small amount of accuracy, but ultimately gives you better gear rankings than trying to guess at which subset of gear combos to simulate more accurately.
(For reference, it takes my fast but not amazing computer about 24 hours to simulate 100,000 combinations. So to simulate 1 billion combinations (which is not unusual even for just a player's bags), it would take 10,000 days.)
So rather than make the user guess which small number of combinations are "probably" the best, then simulate them all and wait hours... we use an advanced algorithm that can approximate very closely in a couple seconds. Could you maybe squeeze out another half a percent or 1% DPS with some hand tweaking and simulation after the fact? Sure, sometimes. But you'll never gain a whole lot, and it got you to a near-perfect solution almost instantly.