That's why the 2 x GTXx80 *always* comes out a touch quicker than the single card dual GPU solution. It has been that way since the days of the GTX280.
Just take a look at the link I posted on the last page (#34). The GTX 690 was a touch quicker than the 2x 680 setup....
I think you have a valid thought process, but this generation we are far from saturating the PCIe lanes in most situations. So the PLX in the 690 works to its advantage - *slightly*.