I have wondered about this as well. The first question that comes to mind is "what is the actual angle of the loco wheels?" to the rail. Are they actually flat/horizontal to the rail? Some older locos have a slight camber, along the lines of the "fast angle wheel" design that helps to center rolling stock on the rails. If locos have an angle to their wheel, then a flat top will not maximize surface contact, a matching tilt/camber to the rail head would.
On my MTH RailKing SD70 the loco wheels appear to be exactly horizontal or flat to the rail (as far as my eyes can tell), so a flat top would maximize surface contact area for this loco.
My LionChief GP also seems to have drivers that are horizontal to the rail.
If this is the norm, then yes a flat rail top will maximize surface area, and will be one factor in maximizing traction.
The number of traction tires on a loco likely outweighs the effect from a flat railhead. Also assuming that magnatraction is not involved, which would make the composition of the rail itself an important factor.
You might try an experiment yourself with your locos with a 40" section of straight Atlas track tilted up at an angle, and measure the tractive effort at the coupler, and compare with your current track at the same angle. If the traction effect is significant, you will be able to tell the difference without complicated devises to measure the force.
Perhaps someone will have some data to share, this is an interesting question to me. My next layout may have a 3% grade on a curve.