I think it would be base layer is your footage, the wipe matte sits on top of that, or rather it is what you’d put in the mask track of each of the video sources with one of them inverted. The matte for the sword (I assume the sword itself doesn’t have a key signal?) on top of that (or dump the video tracks into a container and put the sword matte into container’s mask track) the sparks on top of that and the sword itself sitting on top of everything?