What I can see from a quick glance over that paper:
The author claims that 3 wattmeters connected without Neutral give the same result as 3 wattmeters with artificial Neutral. Just a little difference in final result calculation, for the latter case you just add the three readings (of real power), for the first case you'll have to add them too, but multiply the result by 2/3. This results from each wattmeter sensing phase-to-phase voltage instead of phase-to-(artificial)neutral voltage, that's plausible so far.
For measuring 4 wire unbalanced loads, the correct method would be to connect each wattmeter voltage sense between the phase and the neutral. An artificial neutral is a workaround if the neutral isn't available, the described method saves the artificial neutral (if it is done in hardware). As far as I could see, nothing was said about a offset or "shifted" real neutral (that would differ from the artificial neutral), resulting in different phase to (real) neutral voltages. With the artificial neutral different from the real neutral, the results of both methods would be questionable in the first place.
There's a lot of different power line systems in the world (from an European view the weirdest to be seen in the US of America) - but there's one simple rule: for correct power measurents under all conditions you need one wattmeter less than the wires and all wires connected to the correct arrangement of meters. Four wire installations give best results with Wye connection scheme including the real neutral. For three wire Delta configurations without a neutral, two meters do the job (Hint: Aron connection). For active (real) power, add the results of all involved meters.
The author mixes the topic a bit with sampling, so the reference to averaging the results over a whole number of power cycles - usually you don't have to care about that if you just use a metering chip outputting results like "Active (real) power" and so on.