The 9% number comes from dividing by the number of IPv4 hosts reported by Censys, who do a portscan of the entire IPv4 space.
But obviously most clients will not have any ports open, and wouldn't be visible to the scan. It's not at all correct to treat that as the number of actively used IPv4 addresses.
Right, it's not even close to 9% of all IPv4 hosts.
With that method, it would be more honest to only include the IP addresses of hosts reported by Censys in the numerator as well as the denominator.