When copying file using dd|nc all 4 CPU's (2 cores + hyper threading) are under almost equal load of 40-50% (as shown by top -P).You have a dual core cpu with hyperthreading (enabled or disabled?). Let's asume hyperthreading is enabled (the default on your board), then you have 2 threads per core, so a total of 4 threads, each capable of using 25% of the cpu.
If the process used to handle your transfer uses 2 threads, it would be able to use a maximum of 50% of your cpu. Taking some overhead into acount when reading from disks, syncing the threads etc., the 40% seems reasonable to me.
Running second copy process raise CPU load by no more then 10% saturating gigabit network.
net.inet.tcp.recvspace=131072What settings did you change for the tcp window size, maybe this can solve my dolphin issue.