Unless you have the computer outside somewhere really cold, those temps are incorrect, as others have pointed out. Use Coretemp to check the temps.
I personally watercool because of the large GPU temp drops (reduced by 20-30C usually for me compared to air cooling), and quietness of the whole comp.
I give a nod to all your observations. On the noise issue, that's pretty much an enthusiast assumption about other enthusiasts. Of course, if you use bigger radiators, more fans or doubled number of fans in push-pull configurations, then there is potential for higher noise. And you add some pump noise into the mix.
Water-cooling is more efficient -- or more effective -- than air-cooling. It makes a lot of sense for cooling graphics cards in the mix with a CPU.
On the other hand, the power requirements and temperatures of graphics cards using the newer Maxwell chips are lower.
Now looking back at the CPU aspects, there is a range of temperatures provided by water-cooled solutions such that you're probably not generating electrical noise with higher temperatures, and you might primarily reap the benefits only if you want to push voltages higher. It is possible that you don't need or want these benefits if you want to keep voltages at or below a certain level.
Meanwhile, I've seen here some experiments for making single-fan AiO coolers much more effective than what shows in their ratings. I can also see there are a few air coolers that might give those solutions a "run for the money" with some low-tech enhancement.
The main issue may have been raised by the different thermal profile of some processors versus others. A 4790K may clock to 4.6 Ghz and dissipate less than 100W of thermal energy, while an older processor might be clocked to generate 140W of thermal energy. Yet the temperatures on the Haswell chip are higher or the same with a given cooling solution as compared to the chip dissipating a higher wattage. The smaller die-size and lithography of the Haswell means less heat transferred per unit time across a smaller areal size of the die.
So this presents a sort of "decision threshold" for builders and enthusiasts, leaving them to choose heatpipes, AiO water-cooling, or custom water-cooling. The die-shrink of newer processors moves the bar of a cooling requirement closer to AiO and custom-water standards. Heatpipe coolers are less likely to keep up as lithography and die-size decrease. AiO's may improve, but they will also cost more. Their effectiveness can only be improved by users deploying certain case, airflow and noise-reduction strategies. The same approaches may enhance custom-water solutions.