I have had the exact opposite experience (and we did the test blind). The transport used was the top of the line ($70K) dCS as well as the rest of the dCS stack. The server was connected to the dCS DAC. EVERY time, we would pick the server as the better sounding source (there were two of us who participated in this test). I did the same test on a lower end system with a lower end DAC and got identical results. The server we used was the Sound Science Music Vault.
Something interesting in a tired thread. Can you give us more details on this experience? The server was connected directly to the DAC or to the upsampler? I have to say that I am not interested by the ranking but in the sound difference between the two configurations.