Executing hhmake using FASTA/A3M alignments generate distinct profiles

[apcamargo]

I usually convert my FASTA alignments to A3M using reformat.pl with -M 50 and then build the HMM profiles from them. However, I noticed that if I use hhmake with -M 50 directly on the FASTA file, the resulting HMM will be distinct:

reformat.pl -M 50 fas a3m alignment.afa alignment.a3m
hhmake -i alignment.a3m -o profile_1.hhm
hhmake -i alignment.afa -o profile_2.hhm -M 50

Profile 1:

HHsearch 1.5
NAME  family_100_2
FAM
COM   hhmake -i alignment.a3m -o profile_1.hhm
DATE  Sat May  1 09:31:53 2021
LENG  410 match states, 410 columns in multiple alignment

FILT  139 out of 466 sequences passed filter (-id 90 -cov 0 -qid 0 -qsc -20 -diff 100)
NEFF  8.1

Profile 2:

HHsearch 1.5
NAME  family_100_2
FAM
COM   hhmake -i alignment.afa -o profile_2.hhm -M 50
DATE  Sat May  1 09:35:50 2021
LENG  538 match states, 2200 columns in multiple alignment

FILT  150 out of 466 sequences passed filter (-id 90 -cov 0 -qid 0 -qsc -20 -diff 100)
NEFF  6.9

Is this the expected behavior?