That can't be saying much. This paper blows.
Consider the data from a run taken on 18 Jan. shown in various aspects on page 16, 17 and 18.
At about 2045, alumina powder was added to the setup to increase insulation around the test article. Subsequent behavior is described as "At a constant power of 160 W the temperature increased from 600°C to 1000°C. After this the reactor worked for 38 minutes at a temperature near to 1080°C. "
Now actually look at the data presented on page 16. The power jumps between about 150 and 220 watts. This is clearly caused by the temperature control loop switching heater voltage between two taps of the power transformer as shown in the diagram on page 11. This is hardly "constant power". And the description fails to mention that the second, 38-minute period at 1080 is using an input power of ~220 watts.
OK, that's sloppy. Now look at the analysis on page 17. Input power is claimed to be 144 watts, which is not even the power claimed in the description of the previous, lower-power interval. And computing the total energy of 144 watts for 38 minutes gives 328 kJ, not the 323 listed. And using the 220 watts indicated on the data trace gives a total input energy of 502 kJ, for an output/input ratio of 1.11, which is probably (given the sources of error possible in the setup) about as close to 1 as you'd want.
And then, let's go to the first summary table on page 17. There we see the last entry, with a date of 18 Jan, a duration of 38 minutes, and a temperature of 1080. We also see a claimed input power of 78 watts and an output of 135 watts, with a ratio of 1.73. And you'll notice that neither the input nor the output powers match the previous numbers.
So what, exactly, did you learn from this paper?