I made this argument in a different comment [0], but often the IP address only points to a router. The router then routes the traffic to devices on an opaque LAN (opaque from the WAN side).
In residential settings, there's a collation that whoever owns the router is likely to own the devices on the LAN - but there are also enough cases where this isn't true. Hotels, universities, cafes. Also, Tor exit nodes or people with compromised Wifi. I'm sure it's possible to come up with more scenarios. As for a statistical probability, no idea. Is 80% or 95% acceptable? I think that's EFF's argument, that there's still enough room for error.
Seems I remember a previous article estimating 1.5 million Tor exit nodes, but I may be wrong. And in another article from memory, active ip address estimates were in the 100's of billions maybe. My bet is most illicit activity from a perpetrator would go through Tor, but I wonder how many of the police raids are due to people being 'swatted'?
No, it's not - one example the article lists is a Tor exit node.