No one knows, ultimately, what recidivism is. For one thing, we don't know until the person dies whether they recidivate. So you're quite right; we don't know about offences for which there has never been a conviction. All we can use this data for is a proxy to compare. Is the recidivism rate for people who have been through this process better or worse than those who have been through some other process? I wouldn't want to suggest that any number that's used is an absolute measure of lifetime recidivism.
On September 28th, 2006. See this statement in context.