ALTGX transceiver operates in Lock-to-Reference mode?

If you put the transceiver in lock-to-reference mode, it will not attempt to recover and lock to the clock from the incoming data stream. So essentially your input data will be asynchronous to your clock. I'm not sure what your reason for doing this would be. Typically it's done when oversampling the input data in which case you are responsible for recovering the data in logic.