OK, A couple of thoughts:

- Is the first PIC, on the TX side, going to vary the pulse width of the serial data, to cause the variance for the RX pic to differentiate between the High/Low on channel 4?
- If so, wouldn't that affect the frequency of the serial data, and prevent the RX pic from getting the correct data?

- Why not go with a customized serial protocol? A little more work on the software side, but may be better for this app.

just an idea, but here is something off the top of my head.
Code:
_________|---|===|===|===|===|....|===|___________
   A       B   C0  C1  C2  C3       Cn

A = Specific minimum time for a low in order to differeniante packets.
    Slave watches and when that min time is achieved, 
    it gets ready for the start bit.
B = Start bit, always high, for a specific time.
C = Data bits, one for each line out.  
    C0 is always the status of your fourth channel.
Just some musings as I'm eating my lunch.

Steve