The 10x would be 9x if the methodology were the same. 9x->3x is going from reported accidents to inferred true accident rate, as the article points out.