This looks like a homework question. So you need to face our standard response: "What have you tried?", and otherwise motivation from others to write the complete script for you might be limited.

If you are an absolute beginner in programming, then this task might be a little over your head. I'd recommend to break down the task into smaller sub-tasks, which again I can give in pseudo code. Note that I do not know python, but this should help you to structure the task anyway. That is at least how I normally approach it (and that normally works for me quite well to structure my short scripts, and to spot the need for clarification in the definition ;) )

read genbank file 
declare count_vector := (0,...,0);
declare transitive_set_set_list := [{{}}]; # list of empty set of sets
   foreach pair of sequences {A,B}  in G (genbank file) do:
      run pairwise (reciprocal?) alignment  to get the alignment (I am not sure if that should be part of the task, but I guess so);           
      foreach set of filtering criteria i in 1:10 do:
         if alignment better than filtering criteria
             increment critera_count vector[i]
             transitive_set_set_list[i] := extend_transitive_sets({A,B}, transitive_set_set_list[i]); # this call handles the case of transitive extension
             # this is a bit more complicated
print count_vector
print transitive_set_set_list;

### note that this is almost python code thanks to pythons set datatype
subroutine extend_transitive_sets ({A,B}:set, S:set of sets ) returns set of sets
    locate sets Si1,2 in S: with A in Si1 and B in Si2                    
    Sreturn := {A,B} union Si1 union Si2 # extended sets, there should be max. 2 such nonempty sets, otherwise it's an error
    return (S setdiff {Si1,Si2} ) union Sreturn;

This uses a different approach but I think it should work...

1) Parse a GenBank file, and extract all the protein sequences. Save the sequences as a FastA file ([http://biopython.org/DIST/docs/tutorial/Tutorial.html#htoc12][1]).
2) Convert the fastA file to a BLAST database
3) Blast every protein in the fastA file against the custom database ([here][2])
4) Now parse the BLAST report (xml file), so that the match(es) fits your requirements (% sequence coverage and % identity of at least 10%)
5) format your output