Since CA will take up the delta camp, I guess I'lll represent cartesian. Both are good choices, but if you need it to be easy to troubleshoot and calibrate cartesian is a winner for me. If you need lots of Z height, deltas are an easier choice to implement. However, if you want lots of XY room then I'd go cartesian, the diagonal rods on a delta get a bit wobbly as they get longer (for more XY room). Deltas also get tell pretty fast, both increasing your XY or Z print areas requires a taller machine and you may not have room for this.

If you do go for a makerbot style cartesian printer with a moving bed, your idea of two leadscrews is great, but do not support it at four corners. Three guide rods at most, any more and you are over constraining and it will be more likely to bind up (and I'm guessing you'd had enough of that already with this DLP?). Two guide rods should be enough, too, as long as you position them at opposite ends of the bed so it's not cantilevered. If you take this last route, make sure you have a driving lead screw on each side or it will also probably bind up.

Depending on how (relatively) small or large you want to go, consider different motion systems. On a relatively small (desktop) machine where the motors would make up the majority of the moving weight on the gantry, try making them stationary like in coreXY or sli3DR. If you're going BIG, and the motors would make up just a small portion of the gantry's weight, don't bother trying to make them stationary. Maybe even go for rack and pinion drive like on some CNC mills. You'll get negligible increase in moving mass in exchange for not having to deal with belt stretch or routing.