This sounds like one of those papers that got published with just barely "significant statistical evidence" but never had follow-up to verify, meaning you can't really draw conclusions for it.

The study was barely scientific. The researcher was his own test subject, and the result has never been replicated. Also, most crashes probably occur under conditions where the driver can't be aware of whether the cyclist is wearing a helmet or not. "The cyclist suddenly came out of nowhere" is a common defense.