MetaPhlAn3 results not giving expected profiles

Hi @fbeghini

I have done metaphlan3 analysis for

  1. samples trimmed at 35bp
  2. samples trimmed at 70 bp
  3. samples trimmed at 70 bp and subsampled to a particular read number for each fastq file(sample)

Doubt number 1:
I observed that for 1 and 2 profiles are not the same, but metaphlan3 says it automatically deletes the reads with <70bp read length so, theoretically results for both datasets which are trimmed at 35bp as well as for which the reads are trimmed at 70 bp should give the same profiles. Can you explain this?

Doubt number 2:
When I subsampled all the reads which were trimmed at 70bp the amount of data decreased in all the samples but what I observed is that for more than half of the samples the number of bacterial species profiled by metaphlan3 in subsampled files has increased which seems weird. Why is this happening?

Thanks in Advance