I've had statistically significant greater "catch" factor ( p >=0.001) with traps larger than a 5 frame nuc.
As a matter of course, I make bait hives out of two nucs stacked double. I now either use a medium depth nuc on top or two deeps (frames only in upper nuc). I use 1 or 2 old brood frames and other frames are variously plastic or wired foundationless. No statistical difference in the medium+ deep vs. deep + deep combo. No statistical difference between open frames, plastic frames or number of old brood frames.
Double nucs are a pain to secure --- a scab of wood diagonally, a strap with a binding tension, or hive staples are variously used. The increased size is definitely worth it, however due the much better catch. Swarms usually congest in the upper box, and the empty bottom box can be stripped off when the swarm is retrieved.
Most of my trap boxes are simply placed on the ground. I am in California chaparral and coastal sage scrub, so the bees are used to low habitat. No statistical difference in position of box (ground vs. tree crotch). Cannot really test it (too many variables), but good trapping near the dripline of an isolated tree (theory that this makes navigation to the bait easy).