Yeah, you're not going to get a exact same comparison between chips and power numbers. Some people just win the silicon lottery. You can have the exact same card as someone else and the power consumption can be different.
Chips aren't all the same. When they come out of the foundries they'll require a different amount of voltage due to defects in the process itself.
While you can't compare power consumption between different cards, you can compare frequency. Which is why you ask what frequency someone is running at. You wont be able to reduce your power consumption without reducing frequency as well all else being equal.
These kind of variations are due the wafer positioning, nothing new (and well known, by cpu overclockers). In terms of efficiency for example on lbry the 1070 and 1080, in an overclocking environment or stock frequencies, are almost equal (a bit better in the latter), but if you lower the clocks on the 1080, this chip isn't capable of delivering better MH/W ratios than the 1070. I don't have yet analyzed the 1060 @192bit and @256, but I assume that the ratio it's worse than the 1070, especially considering the undervolting capabilities (86W with 200MH if I remember correctly, with the tpruvot 1.7.8 miner, in my platform - check the link in my sig and you'll find the details).
A true king of efficiency
With the laterst kernel revisions I suppose that we can breach 240 MH/s with the same settings
Gtx 1080 1.8.3, inno3D twinx2, maximum stable hashrate it's 395MH/s (i'll post you a confirmation soon with the settings)