In this specific case, it's because iPad's never got the M3. They're literally comparing it with the previous model of iPad.
There were some disingenuous comparisons throughout the presentation going back to A11 for the first Neural Engine and some comparisons to M1, but the M2 comparison actually makes sense.