Maybe a nit pick. But this isn't a basis to say you can't compare the code. The 'average' code is going to be somewhere between your two extremes. Assuming on average, the code was written by the average programmer, you can get an insight in to what the average programmer of a programming language should expect.
Now it may be that populations of programmers favour different things (speed, memory usage, ease of implementation) but that still forms a valid comparison.
Some of the implementations are extremely optimized and took a lot of effort. Some implementations are not. So you might be comparing highly optimized JavaScript code with naive, below-average TypeScript code. You cannot compare those.
It would be much better if they used the same level of optimization for each language, but they didn't. Furthermore it is called a "Game" (it used to be called a "Shoot-out") because you shouldn't take it seriously. So it shouldn't be the basis of serious research.
What is measured in that table is neither naive code nor the absolute limit you can reach with each language, which means you can't really then compare languages between themselves the way the paper implies.
I think picking professionals at random that practice those languages and ask them to write Computer Language Benchmark Games code would be maybe a bit more representative, but even there you face huge biases.