1. Attribution is not about proofing something, or finding "the one proof", but about verbalizing that there are indications that someone might be responsible. And quantifying this likelihood.
2. Having a reasonable level of attribution of cyber attacks cant ever be realistic if your opponent isnt a moron, as the indications you work with can be artificially created to fool you. There is no way to differentiate between a high likelihood attribution and a clever enough framing. While this is also true for real life attribution, for lets say chemical weapons, it is much easier to create indications in cyber attacks. And the incentives to do so are clear.
You are taking one reasonable critique, that you cant expect a "one proof" in attribution and turn it around to make attribution to something that could actually proof something. The limit of attributions is that you can vocalize, to what degree something looks like it was done by someone. Nothing less but also nothing more.