Your analysis is suffering from survivorship bias -- Italy still has no idea how many infections they actually have.

South Korea is a better dataset because they have tested so many people.