Confusion regarding RFD3 inputs

Hi! I was playing around the input options for RFD3. I think that it is not entirely clear what could the outcome of some combinations mean.

I have a particular example. Consider this input:

{
    "test_2": {
        "input": "./init_pdbs/start.pdb",
        "contig": "A1-263",
        "select_unfixed_sequence": "A69-76,A153-154,A172,A174,A202,A204,A207,A208",
        "select_fixed_atoms": {
            "A1-263": ""
        },
        "length": "263-263"
    }
}

What I expect it to do: diffuse everything from scratch while preserving the sequence in all positions except for ones I specified. However, I see that, for example, Arg92 clearly turns into Phe, despite its number not specified in "select_unfixed_sequence". It is still called "Arg" in the file though. Please see a figure:

How should I specify the input to achieve what I stated - to diffuse everything from scratch while preserving the sequence in all positions except for ones I specified?

Start by tracing how RFD3 interprets the provided input block, especially contig, select_unfixed_sequence, select_fixed_atoms, and length, using ./init_pdbs/start.pdb as the concrete input. Clarify which residues are diffused or sequence-preserved, explain the observed Arg92/Phe result, and document the configuration that produces the requested behavior.