For one, correlation is not causation. For another, these numbers ("3.5% of the whole population", "53% of the time compared to 26% for the violent protests")

For one, correlation is not causation. For another, these numbers ("3.5% of the whole population", "53% of the time compared to 26% for the violent protests") all seems incredibly precise for something as messy as real world protests. Estimating participation, estimating population, estimating "peak engagement - so many components of this are judgement calls at best. And that is before even attempting to engage with cultural and temporal factors. To make such a bold and outlandish claim based on such shaky data feels fraught.