Yes. They should be calibrated regulary for repeatability.
My impression is that accuracy between different types of dynos gets into some fuzzy areas because they don't all measure the same things in the same way. For extreme example with chassis dynos, the roller types calculate power from assumptions about momentum (increasing speed to overcome the mass, etc). How is that accurately compared to an eddy current? How about ones where strain gages are used and force is calculated from that, then torque then power?
I've noticed engine dynos can be run at different rates of increasing rpm. That plays a role too.
Tricky problem getting them to compare to each other.