You are correct that he is skipping all of the RS485 communication between the PC and the axis controller. In fact, in the case of linux cnc the axis controller is effectively eliminated ( it basically becomes a breakout board) as with Linuxcnc all of the control signals originate from the PC.
Our machines use rs485 in several places however. Looking at the PC Mill 50 electrical schematic posted in the files section of this site ( i am looking at page 16, "Axiscontroll x axis"), it shows that most of the signals from the axis controller board to the stepper controller board ( and the spindle control board for that matter) are RS485 signals as well. This is verified in Rich Jones posts.
Honestly, I think this is very good news for your project :)
What this means is that the hardware side decoding of the rs485 communication is already figured out- the Texas Instruments AM26LS31(receiver) and AM26LS32 (transmitter) chips are already documented to work at decoding the RS485 signal to normal ( sig and ground) signals. Emco seemed to really like these chips, as they use the same ones everywhere rs485 is used.
The trick from there is going to be figuring out the actual communication protocol. Rs485 is a method of transmitting a signal over long distances, but it doesnt always describe the actual protocol. You may get very lucky and find out it is RS232 (serial) over RS485 ( which I suspect it may be).
In your position, my next step would be to disassemble the keyboard enough to get good pictures of any circuit boards you may find. What you find here ( or dont find!) Will help us to figure out what we are up against in terms of communicating with this thing.
Keep us posted!