It's a scam.
If this is the actual paper that is supposed to come out in December I can see why it wasn't published in a physics journal. There are a plethora of things wrong with it. So let's start.
In part B they claim a TM212 mode but I'm not exactly sure how they know how to deduce that and how they know how to tune to that mode. Even in their section about tuning they describe how they think the are in resonance but this doesn't mean they know if they are in some particular mode. I'm not an expert in cavities but it seems to be they should have consulted someone who is. They then claim that there are no analytical solutions for a truncated cone, which is not true at all, see here. So right off the bat their understanding of cavities is called into question. They also don't say if their frustum inside is a vacuum, which I think is important if you're going to set up an electric field inside.
They say they put the RF amp on the torsion arm itself. This doesn't seem like a wise choice if they want to reduce all possible systematics.
In their vacuum campaign section they discuss simulated thermal effects but don't say what they used for this simulation. What model did they use, what assumptions were there, etc. If there is a standard piece of software they don't say this either.
In their force measurement procedure section they have a very convoluted and confusing way of measuring force which I don't think matches with their earlier model. One simple way they could have done it is take data with their optical setup then fit it with their earlier thermal model. If they got something significantly above their background model then they might be able to say more. But what they seem to do is record some time series data, what look like pulses, and fit parts of it to linear models to find different parts of some pulse they are looking for. That is a very undergraduate way to do this. They are - from my reading of this confusing method - simply fitting different parts of a pulse to determine what part of the pulse describes a calibration versus other pulses from something else, like a purported thrust. There exists technology that was developed in the 1980s that allows you do do these measurements much easier than they are doing, with much cleaner and clearer results, called NIM, but for some reason they are using this dubious method which likely won't give clear discrimination between signals.
Then they describe different configurations and their effects. The only thing I have to say about this is that it's not clear to me they couldn't have moved electronics outside of the testing area. I've worked with high voltage electronics in a very precise and sensitive test setup before an all of our data acquisition and power supply electronics were easily placed outside the test area, using the technology I mentioned before.
After that they describe force measurement uncertainty, which is great because they didn't have that before. They describe the uncertainties on their measurement and calibration devices. That is fine but these constitute random errors, not systematic errors. The only systematics they talk about are the seismic contributions, for which they quote a number without saying how they arrived at it. They say this is controlled by not doing tests on windy days but that doesn't account for everything since seismic activity, especially from the ocean, can occur without the wind. So it's unclear where they get this number from and if it's at all accurate. This is very dubious. They also cannot control for all low frequency vibration with one method either. Different frequency ranges are usually damped out with different methods. They then say their thermal baseline model contributes some uncertainty, which is true, but then they go and give a "conservative value", which strongly implies they pulled this out of a hat and didn't actually analyze anything to arrive at that number. So I call into question that value. Table 1 tabulates measurement (random) errors then adds them. It looks they quadratically add them, which is correct, but if you worked it out then they did some necessary rounding and didn't keep with the rules for significant figures. They classify seismic and thermal errors as measurement errors, but they are not. If seismic and thermal errors give a continuous shift in your measurements then they should be counted as systematic errors. The authors seem to not understand this.
Their force measurements in table 2 don't seem consistent with what you'd expect to see with increasing power. This says to me there are systematics which they did not account for. In this table they assign an uncertainty to the measured valued which is the one previously discussed. If they has taken data properly and did a proper analysis, the result from that analysis (which should including fitting to their earlier described model) would give different uncertainties for each result. This is standard practice and this is why error analyses are usually done at the end of studies, not in the beginning or middle.
After, they attempt to make some null thrust tests in which they attempt to show that if the z-axis (think in cylindrical coordinates) if parallel to the torsion beam it should show no "thrust". The beam clearly is displaced but since they claim it is not "impulsive" that it is not a true "thrust" signal. This is incredibly disingenuous since it is clear from their plot that something happens with the RF is turned on. The whole idea of impulsive signals doesn't seem correct either since it says to me that they turned they RF on, saw what they wanted to see them turned it off right away. For example in figure 13, would that upward going slow continue to infinity? Probably not. But it's not clear from these plots what the real behavior is.
They then to go on to describe sources of error. At first glance this is great, but upon further reading it looks like an error analysis I would have received from one of my undergraduate students. They are all good sources of error but not a single one was quantified or studied in any detail. At best they simply state in a few sentences why this or that is not important but don't actually back it up with any numbers, which would be proper procedure. This is a huge mark against them and this alone should call into doubt all of their results. But...
They did absolutely no controls. A null test and calibration pulses are not controls. A control lacks the factor being tested (NdT's Cosmos explains this very nicely, episode 5 I think). For that to have been done they would have needed to test several different cavity types: no cavity, rectangular cavity, and most importantly they should have tested a regular cylindrical cavity since this is closest to a frustum. Only then should they have done their frustum measurements. Based on this, their poor treatment of systematics, and their lack of a good method to analyze data (there are no statistical tests mentioned throughout), none of their results should be trusted or given much weight.
They finally go into and start talking about quantum mechanics and how different interpretations could apply (QM doesn't apply here). They also talk about debunked crackpot ideas like Stochastic Electrodynamics (SED), and the Quantum Vacuum Plasma which is complete and utter crankery to anyone who has sat in a half semester of quantum field theory.
tl;dr: It's no wonder why they couldn't get this published in a physics journal. Their experimental and data analysis method are at best at the level of an advanced undergraduate, and they have absolutely zero knowledge of any advanced concepts in physics, which they demonstrate in their discussion section at the end.
This paper should absolutely not be taken as evidence of a working emdrive. And so it remains pathological science.