The model is useable, but could be refined: you're looking for significant sets of paths, defined by source, waypoint and detector, where the contributions add to a significant resultant, so predicting a bright point. For other sets of paths, perhaps terminating in a different detector, the contributions don't add to much, predicting a lack of illumination at that detector. And in principle you've to sum these contributions over all possible paths.
There is a very neat way of working with significant groups of paths: working with three paths, as a triplet, is fruitful. More than three adds complexity without adding insight. Where the paths contribute significantly to the resultant, the contributing arrows line up, where there is a very small contribution, they curl up.
In any situation, picking out what's distinctive about the geometry let's you see how to explore the space with a triplet.