ERROR: Stray PME grid charges detected

Dear all,
I have a system protein+ligand in a water box plus Na+ ions to neutralise the system. I'm using amber imput files. I'm trying to minimize and then equilibrate water and ions with simulated annealing but i'm having a lot of problem. the starting point was the minimized system protein+ligand. Then I made the imput files with amber (sander) then I minimized again the system protein+ligand+boxwater+Na with namd for 100000steps, with periodic boundary conditions, PME anf fixed protein backbone.
here you can find the config file:
--------------------------------------------------------------------------
amber              yes
parmfile           M52SKIrespambermin.prmtop
ambercoor          M52SKIrespambermin.prmcrd
readexclusions     off
outputname         1M52SKIrespsimulatedannealing
restartName        1m52SKIrespsimannrest
set temperature    0

# NOTE: Do not set the initial velocity temperature if you
# have also specified a .vel restart file!
temperature         $temperature

# Periodic Boundary conditions
# NOTE: Do not set the periodic cell basis if you have also
# specified an .xsc restart file!
cellBasisVector1    87.    0.   0.
cellBasisVector2     0.   67   0.
cellBasisVector3     0.    0   91.
cellOrigin         -0.34  0.01 0.13
wrapAll             on

# Force-Field Parameters
exclude             scaled1-4
1-4scaling          0.833333
cutoff              12.
switching           on
switchdist          10.
pairlistdist        14.0

# Integrator Parameters
timestep           0.2;# 1 fs/step
numsteps            100000
rigidBonds          all  ;# needed for 2fs steps
rigidTolerance      0.0005
nonbondedFreq       1
fullElectFrequency  2
stepspercycle       10

#PME (for full-system periodic electrostatics)
PME                 yes
PMEGridSizeX        90
PMEGridSizeY        90
PMEGridSizeZ        120


restartfreq         200     ;# 500steps = every 1ps
dcdfreq             100
outputEnergies      100
outputPressure      100
dcdfile             1m52SKIrespsimann.dcd

# Fixed Atoms Constraint (set PDB beta-column to 1)
fixedAtoms          on
fixedAtomsFile      fix.pdb
fixedAtomsCol       B

minimization       on
if {1} {
minimize            100000
reinitvels          $temperature
}
------------------------------------------------------------

 All run ok until the end and I visualized the trajectory on vmd. From the biginning I have protein and ligand shifted at two opposite edge of the water box, almost all outside the waterbox.

Then I wanted to run the simulated annealing steps, using restart files, here you can find config file:
-------------------------------------------
amber              yes
parmfile           M52SKIrespambermin.prmtop
ambercoor          M52SKIrespambermin.prmcrd
readexclusions     off
outputname         1M52SKIrespMDsimulatedannealing
restartName        1m52SKIrespMDsimannrest


# Continuing a job from the restart files
binCoordinates     1m52SKIrespsimannrest.coor
binVelocities      1m52SKIrespsimannrest.vel ;# remove the "temperature" entry if you use this!
extendedSystem     1m52SKIrespsimannrest.xsc
firsttimestep      100000

wrapAll             on


# Force-Field Parameters
exclude             scaled1-4
1-4scaling          0.833333
cutoff              12.
switching           on
switchdist          10.
pairlistdist        14.0

# Integrator Parameters
timestep            1;# 1 fs/step
numsteps            11822000
rigidBonds          all  ;# needed for 2fs steps
rigidTolerance      0.0005
nonbondedFreq       1
fullElectFrequency  2
stepspercycle       10

#PME (for full-system periodic electrostatics)
PME                 yes
PMEGridSizeX        90
PMEGridSizeY        90
PMEGridSizeZ        120

restartfreq         200     ;# 500steps = every 1ps
dcdfreq             100
outputEnergies      100
outputPressure      100
dcdfile             1m52SKIrespMDsimann.dcd

constraints         on
consRef             restrain.pdb
consKFile           restrain.pdb
consKCol            B


reassignFreq        1000
reassignTemp        0
minimize            1000
for { set TEMP 0 } {$TEMP<601} {incr TEMP 1} {
   run 1000
   reassignTemp $TEMP
}
run  500000
for { set TEMP 600 } {$TEMP>=300} {incr TEMP -1} {
   reassignTemp $TEMP
   run 1000
}
run  500000
run 4000000
--------------------------------------------------------------------
at the first step I have all energy values equal to 99999999.9999
after few minimization steps (50) I have warning message:

ERROR: Stray PME grid charges detected:
and:
Warning: Bad global exclusion count, possible error!
Warning: Increasing cutoff during minimization may avoid this.
Warning: Stray PME grid charges ignored!

at the end of minimization steps (1000)
when simulated annealing should start  I have the error:

ERROR: Constraint failure in RATTLE algorithm for atom 3433!
ERROR: Constraint failure; simulation has become unstable.
End of program
 
if I visualize the trajectory file on vmd I can see protein and ligand at two opposite edge of the water box (as before) and some very long bond, not only with hydrogen, but also C-N, or C-O.

How can I solve the problem? should I enlarge the periodic box or the PME dimension starting from the minimization step?
Should I reduce the timestep?

thank you for your help

regards

Roberta