As I believe Mark Twain once said (though he was quoting someone else), "there are lies, damned lies, and statistics".
The stats above mean very little without more information. For example, Remdesivir is recommended by the NIH to treat "high-risk, nonhospitalized patients with mild to moderate COVID-19". One can reasonably assume those were the guidelines used when treating most patients.
For the "unapproved" treatment options, since there are no guidelines, there may well have been many lower risk patients who were treated which would skew the statistics. I don't know this of course but without that kind of demographic info, the stats aren't very meaningful.