compute chunk/atom gives wrong atom count

I wonder why the output of fix of chunk gives me a wrong number of atoms
This is how the curcial parts of my inpute script look like

compute cc1 all chunk/atom bin/1d x lower 10 units box
compute stpa all stress/atom NULL
variable hydro atom -(c_stpa[1]+c_stpa[2]+c_stpa[3])/3

fix 3 all ave/chunk 100 1 100 cc1 v_hydro density/number density/mass temp ave one file press.profile

The output file of the fix ave/chunk “press.profile” then looks like this

Chunk-averaged data for fix 3 and group density/number

Timestep Number-of-chunks Total-count

Chunk Coord1 Ncount v_hydro density/number density/mass temp

0 1714 4.99995e+06
1 -8381.78 0 0 0 0 0
2 -8371.78 0 0 0 0 0
3 -8361.78 0 0 0 0 0
4 -8351.78 0 0 0 0 0

That is, the 3rd column gives the number of atoms within each individual chunk.

Now if add the number of atoms for all chunks for a selected timestep and compare it to the dump-file at the same timestep,
the numbers don’t match. It seems like the compute misses some atoms.
Morover, the 3rd column of the header (of the fix ave/chunk output file) does give the correct number of atoms.

How can this be? I’m not averaging over multiple steps…

best regards,

please see especially points 3 and 9