Paulo,
You are looking at 2 solutions:
Switching to PML:
You could move to a PML but you will need to make sure that your PML is not close enough to the hole (a resonant structure) that the loss introduced by the BC will not alter the optical characteristics.
Your second concern will be that the example in question is using the PBC with the input plan to achieve a plane wave like incidence, you cannot get this with the PML in place so you will need to reconsider your source.
Keeping PBC:
If you move your PBC out to a sufficient distance that the fields from the hole and any “mirrored” holes does not overlap. This way you would be modelling independent holes and your results could be interpreted as those for a single hole.
My Suggestion:
I would go with the PBC unless you want to model an actual experimental system in which case I would likely go with the PML and use an input configured to match my experimental source.