The way I look at it, the whole deal with lowering variance with multitabling is like this:
Say you play one table for a session and the only big hand you play is AA v KK all in preflop for $10. Now whether you win or lose this session is going to be significant variance. 20% of the time you're supposed to lose all your money here, so if you hold, that's variance because you're above or below the play's EV.
Now instead you play a five tabling session and the same thing happens. At each of the five tables, you get AA in v KK, you're more likely to actually lose one and win the other four, or at least something closer to EV of the sum of those five plays than the result of that single hand in the first example.