The reason given in the paper was that the top ring was wetted less with multivis than with straight grade. Now before anyone gets excited, let me explain ring "wetting."
Visualize the piston and cylinder on its side, with the ring just a few microns from the cylinder wall. Zoom in to just the top ring and the cylinder wall. Assume the end of the ring has a spherical shape to it where it contacts the cylinder wall.
As the piston moves up lets say on compression, or down from combustion, there is a wall of oil on both sides of the ring. The height of this wall is called the "wetting" height. For straight grades, this wall is higher, thus the ring is moving a larger volume of oil and moving against a high wall of oil heats it up locally, and you have viscous drag, thus robbing the piston of some HP (albeit very small). This presents a higher level of friction - viscous friction. (Any resistance while in motion is kinetic friction).
Now the multivis still provides plenting of wetting for the ring, but the wall or "wetting" height is lower, thus less oil is moved, and we have less viscous friction.