SH-aLRT changes when using jackknife

[digitalwright]

I noticed an unexpected discrepancy in SH-aLRT values when using jackknife versus bootstrap.

Attached is an example input file and output trees using commands:

iqtree3  -s M100_DNA_1.fas.gz --ufjack 1000 -alrt 1000 -T AUTO -ntmax 16 --seed 123 --seqtype DNA -m TEST
iqtree3  -s M100_DNA_1.fas.gz -bb 1000 -alrt 1000 -T AUTO -ntmax 16 --seed 123 --seqtype DNA -m TEST

Note the only difference between the two commands is the use of UFBoot or UFJack. Yet, the SH-aLRT values in the output trees are very different, and I do not know why that would be the case.

Further tests suggest SH-aLRT is providing lower resolution values when using any variant of jackknife (-J/-j) relative to any variant of bootstrap (-b/-bb). This occurs across many different input sequence sets.

I wanted to bring this to your attention in case it is unexpected behavior. Thank you for maintaining the code and reviewing this report.

M100_DNA_1.fas.gz
UFBoot.txt
UFJack.txt

[cindykrafft]

PR attached that might help, up to the maintainers to determine if this is a good approach though.