[PyMOL] hydrogen bond detection via the dist command and mode=2 option

Dear PyMOL users and experts,

I wonder if anybody has insight into the following problem concerning 
the determination of hydrogen bonding interactions in PyMOL via the dist 
command with the mode=2 option. For an upcoming local PyMOL workshop, 
which I organize for PhD students, I tried to find a way to detect 
hydrogen bonding interactions of a ligand with its receptor within PyMOL.

This is described to some extent in
https://pymol.org/dokuwiki/doku.php?id=setting:h_bond
or
https://pymolwiki.org/index.php/Displaying_Biochemical_Properties#Hydrogen_bonds_and_Polar_Contacts

The documentation suggests that if resonable positions of the polar 
hydrogens are available, one can detect hydrogen bonds using distance 
and donor angle criteria. I added hydrogen bonds using Protoss of the 
ProteinPlus webserver to the structure 4eiy. However, when analyzing the 
environment of the ligand, I notice that many hydrogen bonds are 
detected quite fine, but there are many others, which are not displayed, 
although they have quite perfect geometry. Furthermore, for some 
hydrogen bonds, the heavy atoms are connected, in contrast to the 
settings (and the correct display of many other interactions).

If somebody wants to test, the pdb file is available here: 
https://research.uni-leipzig.de/straeter/pymol/data/4eiy_protoss.pdb

and these are the commands:

load 4eiy_protoss.pdb
hide all
zoom resn ZMA
clip slab, 8
show sticks
dist hbonds, all, all, mode=2

For example, for water molecules 2522 or 2527 in the vicinity of the 
ligand, the non-hydrogen atoms are connected by the H-bond. Waters 2584, 
2514 and Tyr-9 OH are well set for a hydrogen bonding interaction, but 
this is not detected.

I wonder, what the cause of these apparent inconsistencies might be. Is 
it the algorithm or the detection of the donor/acceptor functionalities?

If the hydrogens are deleted before the running the dist command, these 
hydrogen bonds are shown, but also others which are only very weak based 
on the deviations from donor angle linearity.

Best regards,

Norbert