Yeah, I’ve run into this as well… and come to the conclusion that TL0 just doesn’t work for this. Instead I will sometimes carefully craft some targeted watched words to require approval (or temporarily silence/suspend). I do wish you could individually flag a given account to require approvals independently of TL — I thought that was possible at one time but I can’t find it now.
I used to have a better understanding of this, but my posts about it have all been deleted so 
. I suspect this misunderstanding stems from the Understanding Discourse Trust Levels post — it’d be great if it could get clarified there.