It seems to me ultimately unreasonable to judge three very different systems on the same value scale. At some point, fitness for purpose has to enter into the consideration.
So far, this has been overall the least satisfying SBM cycle I can remember, although I am extremely optimistic that the $350 bonus build may salvage the whole cycle. Others have featured an occasional questionable part, or even a whole build that didn't make much sense, but usually there were some lessons to them; I'm still trying to get the lessons out of this one.
The $650 build is very similar to what I might do on that budget, if building AMD. I would want a more robust motherboard, and an aftermarket cooler, but otherwise it's pretty solid, and has upgrade potential (e.g. a SSD). Whoever wins this one needs to back off the overclock for the sake of the motherboard, but it should otherwise be decent. If I win it, that's what I would do, then probably give it to my nephew (if my sister agrees) or to a friend.
The $1300 machine is very much like what I'd build for myself. I would shift a few things though, like getting a more mainstream mobo and sticking with an air cooler (reducing the OC if necessary), and putting the money into a larger SSD and an additional drive for a RAID1 data pair. If I win this one, it might replace my "Phoenix" rig, but swapping the larger SSD and RAID1 pair from that one. One thing the high-end mobo does provide for this one is a little more future-resistance. I'd expect it to remain upgradable with faster graphics cards for years to come.
I can't think of anyone I've ever known for whom the $2550 machine is anything like what I'd suggest. I'm not sure what I'd do if I won it; maybe I'd part it out, or pull one of the graphics cards to get the power down into a more comfortable range and then sell it.