The complex surface is generated in a confusing way.

[Ieremie]

When computing the iface for parts of the protein complex, which interact with each other, the complex surface should only contain the chains of interest.

For example, Complex_id1 has 5 chains.

We are looking at the interaction between id1_AB and id1_CD parts, however the surface for the Complex_id1 still contains a fifth chain (E). This way we will wrongly label some parts of the id1_AB and id1_BC surfaces as interfaces because they are being covered by an extra chain.

I am assuming we only want to look at interfaces between the parts we know that interact, so I am assuming that using the full complex surface is not the right way to label the interfaces.