Perhaps having the signal passing through the primary before seeing the cap appeals to the purist mindset. But the disadvantage is in applying feedback if you don't want to send DC back to the previous stage or stages without upsetting the bias. In the case of the SP-3, the bias of the 12AX7 is calculated with the DC from the cathode follower. In the more conventional way, it's easier to just feed a resistor back to the input stage's cathode after a DC blocking cap. In a non-feedback circuit, either way is the same theoretically. But having cap on the ground side demands better insulation material on the primary to withstand the higher B+ voltage. Many ways to skin a cat, as the saying goes.
