The amps of the lower one at its operating voltage limits the current the two can deliver in series.
Raw Voc is reasonable to use if the panels are of similar technology, but really it is only for combining to determine if the controller can handle the panels in the configuration you want IMO. If you have two panels, one with Vmpp at 73% of Voc and one at 83%, straight Voc is not too useful in that case for figuring the potential watts you will generate. Vmpp of the system as a whole is what will be used by a perfect MPPT controller.
In this case you might well end up with a Vmpp of the system at about 74% of the overall series Voc because pulling the low percentage panel to any higher voltage would crush the amps and watts the system could put into the controller. If the volts distributed unevenly across the two panels, then all bets are off and you'd have to measure the actual Vmpp to have any idea where it would land.
In parallel it would almost certainly be somewhere around 74% of Voc (assuming the two Voc's are equal) for the same reason - pushing a single panel much past its Vmpp is likely to crush the total watts of the whole system. Thus panels with matched Vmpp at likely operating conditions (NOCT being a decent starting point) matters much more than matched Voc's in my opinion.
Voc is easy to measure though, so if you can find the Vmpp for each panel from charts or by measuring it, then you can likely use the actual current Voc times that percentage of Voc to get Vmpp pretty closely for different measured Voc's on each panel.
Jim