Possible artefacts during trajectory unwrapping using -pbc nojump

GROMACS version:2021
GROMACS modification: No

I am attempting to calculate the lateral diffusion of lipid species from MARTINI2.2 simulations implemented using GROMACS. I followed the following steps for pbc corrections and unwrapping.
Step 1 :- -pbc mol -center ; System for centering and output
Step 2 :- -pbc atom -center ; Membrane for centering and output
Step 3 :- -pbc whole ; Membrane as output
Step 4 :- -pbc nojump ; Membrane output

While calculating the MSD, I observed a significantly non-linear rise in the MSD towards the last 1 microsecond of the simulation. Funnily enough, I observed this for both a trajectory that is 5us long and 10us long! The blow up was only at the last 1us in both cases. Has anyone else encountered such an issue before? I suspect this is some artefact of trajectory unwrapping.
P.S: I used both lipyphilic as well as code that I had written myself to calculate the MSD and got the same results.