'hannob is referring to the reliance of QKD protocols on "an ordinary public communications channel, assumed to be susceptible to eavesdropping but not to the injection or alteration of messages." [0] This means that, unless your authenticated channel itself cannot be broken using quantum computers, using QKD with it will still not let you do secure post-quantum key exchange. If you have a authenticated channel that can't be broken with quantum computers, how did you implement that without either already having or arriving at a shared secret? (If you already have a way of securely arriving at a shared secret, then what do you need QKD for?)

[0]: at page 177 in http://researcher.watson.ibm.com/researcher/files/us-bennetc...