compute gyration in rerun


I am rerunning over some dump files. It gives me very nice values for com (compute com), and interaction energies (compute group/group). However, the results for radius of gyration (Rg) (compute gyration) are not sensible; the initial value for Rg is fine (between 15-20 is good), but it increases constantly (i.e. the trend is strange) (see figure please).
I guess, there should not be any problem with the process of read-dumping, because the values of com and interaction energies are consistent with what is happening in the system (vmd visualizations). However, Rg values don’t.
Do you have any idea what happens with Rg calculations? And what is wrong here? The commands I use are:

dump 1 all custom 5000 dump.lammpstrj id mol type element mass q xu yu zu x y z ix iy iz


compute gr1 polymer gyration
fix gr1 polymer ave/time 5000 1 5000 c_gr1 file

read_dump dump.lammpstrj 0 x y z format native box yes scaled no wrapped yes

rerun dump.lammpstrj first 0 every 0 dump x y z box yes


Can you verify that the Rg values from the rerun are inconsistent

with Rg values computed on the fly (via compute gyration) in

the run that created the dump files? Ditto for COM. Compute com

and compute gyration do the same thing with the image flags,

so I don’t see an obvious reason why one would work and the other not.