I did not go to all these interesting references, but I want to confirm the math part:
26.4%, down to 20.1% would be a 23.9%.
One question is what is uncertainty of the estimates.
You would also need to accept that there might not be a simple answer here: there are multiple variations to be tested: doses, early/late interventions, combinations, and it will take time to go through all of them at high confidence.
The question is how to act under uncertainty. The answer is to balance risk and reward.
I saw some articles that advocates several cocktails with multiple specific components, including hCQ, zinc, etc as long as they are relatively safe. Trying to uses and evaluate a cocktail together, instead of separate elements.