Look at this image (
http://img59.imageshack.us/img59/4673/waypoints7ou.jpg) to illustrate it more visually.
The waypoint walking system, defined in k_inc_walkways, would tell the character Move from Waypoint 1 to Waypoint 2. This in itself would be equivalent of the red arrow in the image.
However, since there is a fence and an obelisk in the way, the character, standing at Waypoint 1, cannot reach Waypoint 2 by walking in a straight line.
This is where the pathfinding network comes in. It defines a number of nodes that are tied together that makes up valid paths NPCs can use to properly navigate the map. In this case, the relevant paths that could be travelled to reach Waypoint 2, as specified in the .pth file in KotOR, could in this case look like the green arrows.
In other words, the script (using k_inc_walkways) tells the game engine Move the NPC from Waypoint 1 to Waypoint 2, and the game engine looks up the paths, as specified in the .pth file, and picks what it deems the shortest route to reach Waypoint 2.
The biggest problem with the pathfinding (aside from it being hideously CPU intensive) it that it does not automatically take obstacles, like placeables, into account. Thus if you edit an existing area and happen to put a placeable on top of a path, NPCs are likely to bump into the placeable and be unable to continue since they can't find their way around it. In cases where the map and "real terrain" diverge, NPCs always trust the map. :)