To me that machine looks like an excellent basis and the changes you suggest sound worthwhile, particularly the triangulation. I would be inclined to leave the sides open as it may be convenient for long pieces. By all means make the steel box section that supports the rails bigger to compensate.

Milling the surfaces for the linear bearings flat is a good idea, but you want to do it after all of the welding as any subsequent welding may throw it off again. If the milling machine is big enough then it's easy enough to clamp the whole frame to the bed - the difficult bit is clamping it to the bed without distorting. If it twists when you clamp it to the bed, then reverts when you take it off then it's a waste of time...

An alternative to milling is using epoxy resin. Put barriers round the top of the box section where the rails round with a channel / tube between. Pour the resin in (couple of mm thick I reckon) and gravity will self level it extremely accurately.

You can mill the spoil board with the router when it's finished. Having an adjustable height bed is great but it's hard to keep it level when you move it!

I can highly recommend the PM752's. I've neither seen nor used the G540 in real life but I doubt the 'clever stuff' it does is worthwhile for a machine this size.

For that length I think you'd be fine with 16mm, 10mm pitch ballscrew, not 20mm diameter. The bigger screws have a higher inertia which will harm your acceleration and top speed to an extent, however the critical speed is greater so with big enough motors it would go faster. I managed to get 60m/min on my 16mm, 10mm pitch Y-axis which is a 910mm long screw, about 840mm unsupported ... which is clearly much more than enough! That is with PM752 and 3Nm Nema 23 motors. You don't want Nema 34 for a machine this size as the moment of inertia of the rotor is greater which, among other things, means they operate at lower rpm.