In comparison with the new ISA-JSON format, the internal data-structure (see for example in this test script) does NOT handle process edges properly.
In Bio-Parser-ISATab, entity nodes have protocols but these belong to the node, not the edge connecting to the child entity node. Therefore it is assumed that all protocols will be applied in the creation of all child nodes. So, you couldn't have, for example, a Source that creates two different Samples via two different protocols.
Apologies for the misunderstanding ten years ago!