Its to do with how the Firstbeat algorithms work and the level of accuracy they currently require.
So the stress score can be calculated with the current Garmin OHRM because it's good enough to pick up a decent HRV signal when you're moving around.
It's not good enough to capture a high fidelity HRV signal when you're running however, which is why you still need to use a chest strap to get performance metrics like lactate threshold.
Because they are just data based algorithms Firstbeat may one day figure out a way to calculate some of the performance metrics that still require a chest strap. It wasn't that long ago that getting a "stress score" required doing a 1 minute test with a chest strap every day.