in #1, you delay MB orbs by 1 GCD, and SW:D orbs by two GCDs
in #2, you delay SW:D orbs by 1 GCD, and MB orbs by two GCDs
in #3, you don't delay SW:D orbs at all, and delay MB by two GCDs.
#3 gives you a net gain in orb production.
This is why it's rarely useful to think of SW:D as anything but a single spell that takes two GCDs to cast, unless you're sniping orbs off dying adds.