I got a helpful guy over at the Nvidia forum to help me with this issue, so updating this for solution and posterity (just in case someone else find themselves in the same situation in the future).
The correct (form) metamode line should read:
Option "metamodes" "DFP-1: 1280x720 { ViewPortOut=1240x680+20+20 }"
Notice that the dimensions attached to ViewPortOut has to be decreased with twice the amount of the right/down shift.
E.g. in my situation: 1280 - 2 x 20 = 1240; and: 720 - 20 - 20 = 680.
(Obvious, once you know, my tired eyes missed the number differences in the README.)
This
does reduce the pixel area by that amount, but retains a pixel-perfect ratio and thus the sharpness.
Some applications might not like this - games, MythTV - in which case a scaling might be a better solution. This is equivalent to the old overscan compensation and is achieved by adding a ViewPortIn to the line, e.g.:
Option "metamodes" "DFP-1: 1280x720 { ViewPortOut=1240x680+20+20, ViewPortIn=1280x720 }"
This gives the full pixel area but is scaled to fit the available screen area, so not pixel-perfect and less sharp for that reason.
Hope this might help someone in the future.
For me, I am settled and shall mark this solved.
Cheers!
CJ