He's getting two grams of force per 15 KG of total apparatus, using some 1200+ watts of input. shielding included, according to the eureka article.
Not terribly efficient, unless a true nuclear reactor is in use. (as opposed to RTG's) He's about one order of magnitude less efficient on thrust than current in-use RTG powered Ion Drives for DS1.
It sounds psuedo-science to me; grams of force as opposed to newtons is, well, unusual. Not inappropriate, especially in the context given where he's measuring the "effective mass" in to different orientations and deriving a net 4g different measurement of weight.
I'm leery, though, since the mass is not evenly distributed, if his scale is being effected by the possible change in Center of gravity.