There is no such list of criteria. It depends on the interpretation of the scientist. Important factors that come into play in protein interaction analysis can also be physical conditions in protein interaction analysis. In other words, one of the two protein structures may have passed into a recess of the other protein structure in the form of a ‘notch’, which we can characterise as a ‘cavity’. This often makes it difficult for the two structures to "let go" of each other. On the other hand, on the other hand, the fact that you get extremely few interactions will mislead you if you ignore the ‘notch’ interpenetration. Many cases like the one I gave as an example make it impossible to set a ‘criterion’. To overcome such problems and to see the continuity of the interactions you obtain, molecular dynamics simulations can be used and the interaction of two protein structures (or chains interacting with each other) can be analysed as a function of time. The real important effects reflecting in vivo conditions are revealed by such studies. However, it should not be forgotten that, although technological advances provide high quality results in terms of closeness to experimental results, the scientific literature requires that experiments be carried out and that information about them be presented. This often makes it difficult for the two structures to "let go" of each other.

The method you use increases your chance of error when done manually and is below the standards accepted in the literature in terms of analysis. The part of your general analysis method that needs to be improved, the part you are currently working on, is not faulty but incomplete. The part you are currently working on, selecting the interactions that are below 5 Å in terms of proximity, is quite logical. Considering that you are doing this manually, I suggest you create Python scripts to eliminate the chance of error and automate the analysis. The MDAnalysis library is very useful for this.

The scientific topic that will benefit you in your studies is "non-covalent interactions". Considering that you already know the main reason for excluding two of the non-covalent interactions, you may consider examining other types of non-covalent interactions. The types of interactions you excluded are the weakest non-covalent interactions. One of the strongest and most important ones is hydrogen bonds. Hydrogen bonds show a proximity of around 3 Å between two atoms. Salt bridges are interactions between two oppositely charged groups. etc. etc. etc. You can eliminate weak interactions with a logical inference by sorting according to the strength of the interaction.