If you rig up a sensitive DTI with a long arm on a mag base clamped to a 'flat' (small surface plate) upside down on the big surface plate, you can extrapolate the flatness of the plate out to maybe 0.7 of the diagonal.
A lot then depends on the sag of the cantilevered overhanging straightedge, which varies with the cube of the depth of the straightedge.
You can evaluate that with the rig I described, by starting with the end of the straightedge just on the plate. Qualify the part of the straightedge that's on the plate with the DTI, then move the straightedge and the DTI progressively off the plate until the self-weight deflection starts to be a problem.
Obviously the DTI is essentially for picking trends rather than evaluating local flatness, but you can check the latter by spotting even if the straightedge is twice as long as the plate, provided you dream up a suitable way of supporting the far end of the straightedge. It's the straightness trends over a distance significantly longer than the diagonal which you can't check this way.
Having said all this, I'm a metrology munchkin as you can probably guess from my post - you'd probably get more sense over on the new metrology forum.