when calculating arc flash levels, is it preferred to use the secondary (IDMT overcurrent) protection relay settings or ar the primary protection relay settings used for the arc interruption times?

It depends.

For an arcing fault that 'involves' the secondary protective device (i.e. in the same enclosure), you would have to use the primary settings. For a fault sufficiently removed from the secondary device (i.e. in a different enclosure) then the secondary settings should be usable.

Using the primary settings will most likely result in higher arc flash energies. This because the primary device is set to protect the primary of the transformer. If a fault occurs at the secondary windings or downstream, then to the primary device, it looks like a big overload (due to the current transformation), and it will take a considerable amount of time (it often exceeds the 2 second point) to trip.

This is why the area between the transformer secondary and the line side of the main service breaker is the most hazardous place to work. For larger transformers, it can be HRC 4 or even Dangerous.

Differential protection can solve this issue by creating a differential zone between the line side of the transformer primary device and the load side of the main breaker. If a fault occurs in this zone, it is quickly removed. Reducing the trip time will lower the arc flash energy.