165W was measured at the socket using a power meter, so that was for the main system unit.
65W is the CPU power consumption measured by HardWare Monitor Pro (Windows .exe link!) presently 63W, and it's very accurate.

Even a modestly overclocked 130W CPU is not likely to be consuming the full 130W, but obviously the more you push it the more power is required. At stock, crunching flat out, I would expect it to be around 90W. I would be surprised to see it much over 100W. With a very modest freq. bump 100W would be my guess (for crunching here). Again I would expect the motherboards to consume about the same power so I will stick with 200W for the main system unit.

Socket measurements include the inefficiencies of the PSU, so it's the only way to get an accurate picture of the overall systems power usage.
However it's often the case that you need to know how much power a component is using; anyone that overclocks and doesn't bother to check the power usages and temps is playing with fire. In my experience modern CPU's (Intel especially) and modern GPU's use well below their TDP when crunching (and I have tested this on many projects). Only stress testers and maths projects reach the TDP, and while there are a few programs that can push a CPU or GPU past it's TDP, these are basically designed to do just that, and should not be used extensively. Some GPU's prevent the overuse of power to safeguard against the misuse of such programs.

I see there are rumors of an i7-2800K (really for those with LGA1155 boards looking to upgrade). This highlights the gap between the TDP and actual Wattage when the CPU is heavily used, but I would have expected this anyway; Intel almost always have an initial release of several CPU's for a given board and then add to the list in the future; i7-980, i7-980X, i7-970, i7-990X.

Many of the motherboard advances since the initial i7-2600 release are reducing the gap between this generation and the 2011 generation; you can now use both the on-die GPU and separate GPU's, and Gen3 PCIE slots have just come out. When you think about it there is not much difference other than dual vs quad channel RAM. I think Intel deliberately nobbled the LGA1155 systems with dual channel (remember the 1366 boards used triple channel) to keep the i7-980 to 990 as the top systems and keep the door open for the next generation of top system (LGA-2011). They did the same with the i7-800 series, basically just made them a bit more energy efficient. Add DDR3 to the 1155 boards and the difference between an i7-2600K system and an i7-990 would be negligible, meaning the 1155 systems would not really have had a place in the market. The same applied to the high end i7-800 systems, which could have then been competing better against the i7-2000 systems. You can't really OC the fsb of these 1155 systems, or they would again narrow the gap to the i7-980/990's.

Don't know where you found your testing results, but when I tested this on my i7-920 I observed improvement when using triple channel verses dual channel and dual channel over single channel. RAM speed also made a difference. It varied by project/task and also impacted on GPU projects to some extent.

Not surprised about the heatsink problems; that has been an issue on many previous boards. Bottom line is that board just does not suit a big fat heasink - completely down to Gigabyte's design team. Look where they put the power button. Just beside the RAM where all the PSU cables will pass over - basically unreachable.