Nothing gets created in the script....it takes what you've placed in the level and uses the skill1 values to keep everything paired up and working right. You can also use any kind of entity that can have an action assigned to it for the camera, so that means models, map entities, or sprites.

I think the only way you can get access to an actual "position" would be to use "str_for_entname"(I think that's it) to get the name of the position(pos_000, pos,001, etc) and get that into the array in the "set_cam" action some how.

I think it would be less hassle to simply replace the positions themselves with an entity and use the scripts as designed. Since you are using MAX2GS, I know you can export non-level block models and sprites with the level itself, or you can do it like myrlyn68 suggested.