Thanks for the offer shed tuner, but it was the buffer as expected. Narrowing down the buffer preload gave increasing fps on the lighter pistons, and almost no change on the heavier pistons.
Moving from 1.25mm to 0.25mm buffer preload gave an extra 50fps. It's quite sensitive but that's to be expected.

Dvd/rustybuzz. Diana giss is a good enough assumption without giving too much away. There are plenty of ways to go about it though. The mass doesn't have to equal the position mass if you gear up or gear down. You can play with ratios of mass and speed and still achieve the recoiless effect. Or the mass could be split into multiple parts. A collar around the outside of the active piston, and a heavy dynamic spring guide, linked together with cable or chain. This way the active piston can pass through the middle of the dummy to get more stroke and make 12 ft-lb.

Backlash is a funny one. With new parts (to spec) the backlash is around 0.3mm which seems to be OK. However if there any abuse loads (for example dry firing or firing without the buffer) the wear it causes on the gear mounting causes some slack. I've got some abused parts that have almost 1mm of backlash which is noticeable now.
I do have an anti-backlash setup just in case but was hoping to to need it. Time will tell!