You were referring to data more than a week old, when the infected cases had neither time to recover nor time to die. The latest figure from yesterday is 1.16% of all cases according to JHU.
> That's not how CFR works
That is exactly how CFR works. If you test more people, C will be closer to I.