The argument goes something like this: In a top-two primary, situations would arise in which districts that are usually closely divided between the Democrats and Republicans could wind up with a general election featuring two candidates from the same party - effectively wiping out the other party's "right" to contest a close seat in the November election. Presumably, that situation could arise with four (or more) candidates tightly bunched together -- say Donnie Democrat 26%, Lucy Liberal 25%, Ralph Republican 24%, Connie Conservative 23%. And while the Democrats combined for only 51% of the vote, they would get 100% of the spots in the general election (acing out the GOP.)
But here's the thing: In Washington's inaugural top-two primary, across 124 separate legislative races, that didn't happen a single time. In fact, in every single swing district, the top two candidates were one Democrat and one Republican.
Read his entire post, its well worth your time.
Also, given the weird situation with Halvorson, Romero and whatever the Republican's name was, this "swing districts will still elect a Republican and a Democrat" theory holds water.