1. You should put the same-sized modules in the same-color sockets. So if your RAM sockets are black-blue-black-blue, you should put them in 4-2-4-2 or 2-4-2-4 for the best performance.
2. If you put it in 4-4-2-2, It will run 2GB out of the 4GB modules in dual channel with the corresponding 2GB modules in the other bank, and the other 2GB will be in single-channel mode. So this is like "partial dual-channel".
In short, matching capacity = put in matching colors. That will give you the best performance. Although it has been proven by Tom's Hardware than single vs. dual channel doesn't really make a huge difference in real-world applications.
I think you're misunderstanding dual-channel mode. Dual channel mode will interleave the memory so that your (theoretical) memory performance is doubled. For example, if one of your modules can do 8GB/s, and the other one can also do 8GB/s, then in dual channel, they can do 16GB/s. They will not, however, decrease the capacity of your total setup. So if you use 2x4GB modules, you will have all the 8GB available to your computer. It will not halve the system's memory capacity.
If you don't really need that much RAM, I suggest you just use the 2x4GB modules. However, if you want to use all of them, 4-2-4-2 will give you the best performance. It will let you use all 12GB of RAM (assuming you have a 64-bit OS) and will run in dual-channel mode. It will run in the speed of the slower module, though, so your 1600MHz 4GB modules will run in 1333MHz only. But in my opinion, the speed you "lose" is worth the increase in RAM if you really need the memory capacity.
Interesting article. Based on those results...basically...faster memory does almost nothing. At the very least, it improves perfromance by a miniscule percentage...ie...in value based computer building, is it correct to say then that I should almost not care about cas and timings and just got for the cheapest reliable ddr3 of 1333/1600 I can find?