"Prove" is a dangerous word. Everyone involved in the testing of this device is someone who wants desperately to see it succeed. When the effects you're measuring on on the order of 50 microNewtons, it doesn't take much of anything to screw up the results. Read about the history of N-Rays for a historical example of how even (or maybe especially) very intelligent, informed people can fool themselves into believing poor experimental results.
Three experiments does not overturn 300 years of experimental evidence in support of conservation of momentum. Extraordinary claims require extraordinary evidence. So far the evidence has been interesting, but not extraordinary. Like the article says, show me an experiment with thrust correlating with power input. Show me another one where the device runs for a month. But most importantly, show me one performed by skeptics!