Expected

• understand the filtering rules of mmseqs filterresult
• understand the output dumped by mmseqs filterresult
• have a basic understand of the format produced by mmseqs filterresult. It looks like alignment format.

Current Behavior

When i executed mmseqs filterresult db db db_clu db_clu_nr --diff 2 to get only the 2 most divergent sequences in each cluster, it seems:

• The unaligned seqs (one cluster has only one seq member) was discarded.
• When the cluster has only 2 seq members, only the not representative seq was kept.
• When the cluster has more than 3 seq members, 1 or 2 seq members which were not representative seq were kept.

Steps to Reproduce (for bugs)

Take the whole protein of human as example

mmseqs createdb human_protein.fasta human_protein_db

mmseqs linclust human_protein_db human_protein_db_clu

mmseqs filterresult human_protein_db human_protein_db human_protein_db_clu human_protein_db_clu_filter --diff 2

Your Environment

Include as many relevant details about the environment you experienced the bug in.

• Git commit used (The string after "MMseqs Version:" when you execute MMseqs without any parameters): 14.7e284
• Operating system and version: Ubuntu 22.04