LINCS error during Alchemical free energy calculation

I want to discuss the issue i am facing with pmx based free energy calculations.
Briefly i have taken my protein which is in bound and unbound conformation, Then using pmx web-server i introduce one mutation (F2L) at position 46 in my protein.
Then i have prepared files for forward and reverse transformation for both bound and unbound conformation.
I did 1. Minimization (max steps 100000), 2. the NPT simulation for 1ns with positional restraints and 3. 100ns equilibrium simulation
After completing this i have split trajectory frames starting from 50ns into 50 input files for non-equilibrium simulation to run them for 5ns.
But during this stage LINCS error start appearing after 85ps of simulation and simulation stopped.

I did several modifications to avoid LINCS error during my simulation,
Ex. 1. Changing soft core potential value from 0.3 to 0.2 or 0.5,
2. I used PME method but also tested Reaction field method,
3. changes in tau_p values,
4. Simulation steps size from 0.002 to 0.001,
5. Constrained option in LINCS from all atoms to H-atoms only
6. Centering the protein during trjconv step,
7. Running the energy minimization on gromacs output file after trjconv step exactly before non-equilibrium simulation.
but none of these options worked for me.

The error I am getting is for the mutated F2L atoms as i confirmed this by looking into the topology files and LINCS error.

During input file preparation i have got two warning which i have ignored by using -maxwarn 2 (One of this warning is about presence of charge 1.0000 in B-state (Do i need to worry about this charge? or this charge is the source of error? How i can avoid this?