Can only say that by breadboarding the PICDemZ schematic I can start a node1 (or at least it says it has started a network). The node 2 device is a different story though, it gets stuck in a "can't find a suitable network" loop. So presumably, the end device never fires an interrupt for some reason. Have had no better success compiling the example code either.
The setup used is SE C18 v3.3, MPLAB 8.3, and recent pre-compiled hex files. Devices used are a 18LF4620 and a 18f4620 with a 74hc244 level converter.
One indication that I haven't killed off the MRF24J40MA modules is by downloading the **broken link removed** code. With that, at least some communication was established between the modules. That could be an angle, because ascii commands are used.