While I don't think many are advocating for a "fantasy land" with magic oracles to give us Perfect Truth, we can certainly approach the truth of guilt or innocence, using the courts and their accreted layers of checks and balances, such as appeals, standards of proof, and revisionable legal doctrines.
And when we reasonably dispense with the requirement of "Perfect Knowledge" we are left with the imperfect but practical and broadly satisfactory result that when the courts have adjudicated someone is guilty, they most likely are. This guilt "beyond reasonable doubt" means the system is able to express and satisfy what you ( I think ) rightly say is the sense of a decent chunk of people that death is appropriate for some crimes ( not saying I agree with it being so ). That we can approach truth "to within some epsilon" seems even more correct regarding the death penalty in the US, where, at the end of death row, you've now been through many appeals ( I think, I'm no expert ).
So shifting things back from your analogy to adjudication of which materials are "extremist" (or whatever language the actual proposed bill uses), I agree ( and I think few would dissent ) that we cannot magically know for sure whether it is or isn't so, but I want to expand your thinking to encompass the idea that we can approach the truth of whether it is or isn't, with a reasonable deal of accuracy.
And given we ought to be able to approach such a judgement reasonably accurately, I feel a system like this will be able to express the notion that: censorship or banning or punishment for propagation or use of some materials and information is a reasonable punishment, because those informations are so harmful.
So, not perfect, but mostly effective, is how I would wrap that up.
A second, more general, point this leads me to is that our systems do not begin as perfect, they evolve, and we must be prepared, as a collective of people, to take risks in order that we have progress.
When new laws and so on are introduced we are not experts in them, and there is fallout, and we make mistakes, and innocent people get hurt, but society makes progress. I think the risks we are prepared to take ought to be proportional to the reduction in harm achievable, and with things like terrorist propaganda, we are talking about significant harm, and the chance to reduce it. So, to risk wrongful imprisonment, is not perfect, and it's a tragedy to innocents, but I think it's an acceptable risk given the consequences of a failure to intervene.
It would certainly be nice to ensure zero harm is ever done, but in practice, we do not have such magic oracles to tell us how to achieve that, so on the road to progress, and getting results that improve us, we fuck up and people get hurt. It's not to condone in any way that this happens, but we must face the reality that this happens, be motivated by that reality to improve those false positives, and recognize that as we make introduce new laws and norms and make efforts to intervene and improve society, there are risks, just like with any new technology.
Ultimately I think society decides whether the risks are worth the potential improvement.