Why are the simulation risult different, between rigid an not rigid the pore wall?

Dear lammps developers and users,
I am using the lammps to investigate the wettability of the water on illite. In my simulation system, the methane and the water are located in the quartz slit pore. I use thed fix nvt command. If I donot rigid the quartz wall, the water will be adsorbed on the quartz surface, which is reasonable. But if I use the setforce 0 0 0 and addforce 0 0 0 command to rigid the quartz wall, the water will not be adsorbed on the quartz surface, which is not reasonable.
Could you please help me to see why dose this happen?
The in file is as follows:

log 1.2H2O8nmCH4Quartz_nvt.log

units real
atom_style full
dimension 3
boundary p p p

LJ potentials

pair_style lj/cut/coul/long 12.5
bond_style harmonic
angle_style harmonic

pair_modify mix arithmetic

kspace_style pppm 1.0e-5

read_restart 1.2H2O8nmCH4Quartz_min.restart

#pair_coeff * * 1.0 1.0 2.5

neighbor 2 bin
neigh_modify every 1 delay 0 check yes

define groups

group lower id 1:7650
group upper id 7651:15300
group methane id 15301:18500
group water id 18501:18680

group flow union methane water

initial velocities

compute mobile all temp
velocity all create 363.15 482748 temp mobile

fix 111 all nvt temp 363.15 363.15 100

#velocity lower set 0.0 0.0 0.0
#fix 1 lower setforce 0.0 0.0 0.0
#fix 2 lower aveforce 0.0 0.0 0.0

#velocity upper set 0.0 0.0 0.0
#fix 3 upper setforce 0.0 0.0 0.0
#fix 4 upper aveforce 0.0 0.0 0

Run

timestep 1
thermo 1000
thermo_style multi
thermo_modify temp mobile

dump 1 all atom 500 1.2H2O8nmCH4Quartz_nvt.lammpstrj

run 50000
#minimize 1.0e-7 1.0e-9 1000 100000

write_data 1.2H2O8nmCH4Quartz_nvt.data
write_restart 1.2H2O8nmCH4Quartz_nvt.restart

Many Thanks,

Guohui Chen

Dear lammps developers and users,