Q: MACS3 hmmratac for single-cell ATAC-seq

[acv21]

Hello,

I would like to use the MACS3 hmmratac tool to peak call my 10x multiome (single-cell ATAC-seq) dataset. I firstly ran the cutoff analysis as follows:
macs3 hmmratac -i file.bam -n file -f BAMPE --outdir ./dir --cutoff-analysis-only

I am attaching two output files, from two separate samples:
sample2_cutoff_analysis.txt
sample1_cutoff_analysis.txt

I have two questions:

1. The documentation suggests that the lower cutoff should capture a moderate number (10k) of normal-sized peaks (500-1000 bp), whilst the upper cutoff should capture typically hundreds of very high fold change, abnormally large peaks. Looking at my outputs, it seems like my lower cutoff should be higher than the upper cutoff. E.g. in sample 1 I would set the lower cutoff around the 600 bp average peak length mark (so -l = 77), and the upper cutoff around the 2000 bp average peak length mark (so -u = 4). Shouldn't these numbers be the opposite way around (-l being smaller than -u)? Which cutoffs would you suggest selecting for my samples?
2. These two samples are part of a group of multiple samples that I would need to analyse together and compare to each other. However, as you can see, the outputs from the cutoff analysis are quite different, and I would pick different thresholds for these two samples. Should I attempt to pick the same cutoffs for both, for downstream analyses e.g. differential accessibility, etc., or is it fine to pick different, sample-specific cutoffs?

Thanks for your time!