I'm not so confident on this bias claim. Biasing high would be an extremely risky decision for a glucose monitoring company, because it encourages insulin dosing which could be deadly.

Biasing low would be the move if scared of risky downsides.

When performing calibrations I typically found my (accurate) blood finger prick monitor to be 1 mmol/L lower.

It’s not enough of a difference to be deadly, as the sibling comment suggests, and the clinician guidance is to always eat something if you are feeling hypo, and at or near that level.

The upshot is that I set my CGM-linked pump to target 1 mmol/L higher to compensate.