What you're describing is a scenario where you take a sample, and use its measure of central distribution to make inferences about SAME METRIC in the population. This is perfectly valid!
But the obvious flaw with your argument isn't "this paper says that 75% of X volume is speculation, but it's excluding some data", it's that the paper presents literally no claims or evidence about the % of volume that is speculation lmao.
You have been conflating 2 entirely unrelated metrics this whole time. I said this before, but I guess it needs repeating: the % of volume that happens on-exchange is not a proxy for the % of volume that is speculation. These are completely different things!
If you want an accurate analogy, this would be like if you said "this paper shows that the average human lifespan is 4 years", but what the paper actually said was that the Olympics are held every 4 years. It's a completely unrelated data point, irrelevant to the topic at hand, and you'd have to either be ignorant or intentionally deceitful to pretend that they are the same thing.
Just bush league stuff... and as I said earlier, if you can't grasp this concept then I don't know what you're doing on a crypto thread.
I actually do think that you have serious underlying confusion about the nature of what speculation even is, because you keep pretending like it's something which can be objectively measured by using on-chain data. Data is obviously a useful tool when it's available, but making up bullshit statistics to support your argument is malignant.