about Dimer calculations

[dwyuan2010]

hello everybody !
I am calculating CO oxidation on metal surface with dimer method.
After converged, I obained this results in DIMCAR as following:
Step Force Torque Energy Curvature Angle
1 0.21493 6.76320 -138.77518 -6.02919 5.67156
1 0.21493 4.71126 -138.77518 -6.89996 7.23314
1 0.21493 4.74963 -138.77518 -7.13029 7.04969
1 0.21493 2.41667 -138.77518 -7.58621 1.77555
2 0.49640 2.05426 -138.77620 -7.20335 1.35661
2 0.49640 1.55708 -138.77620 -7.26379 1.64673
2 0.49640 1.63354 -138.77620 -7.28929 0.92251
2 0.49640 1.05582 -138.77620 -7.32232 1.31843
3 0.16505 2.01933 -138.78061 -7.70867 2.01034
3 0.16505 2.18501 -138.78061 -7.79636 1.62863
3 0.16505 1.45221 -138.78061 -7.83888 0.95182
3 0.16505 1.17448 -138.78061 -7.87855 1.18470
4 0.26086 1.49509 -138.78154 -7.88175 1.02202
4 0.26086 1.16380 -138.78154 -7.94741 1.04888
4 0.26086 1.11711 -138.78154 -8.01305 0.63668
4 0.26086 0.71881 -138.78154 -7.94784 0.82465
5 0.22743 1.42895 -138.78262 -8.11679 1.12131
5 0.22743 1.22380 -138.78262 -8.15650 1.03900
5 0.22743 1.02606 -138.78262 -8.16399 0.69025
5 0.22743 0.78474 -138.78262 -8.19248 0.73382
6 0.18357 1.81808 -138.78316 -8.25363 1.03289
6 0.18357 1.04579 -138.78316 -8.29907 1.24204
6 0.18357 1.14508 -138.78316 -8.39061 0.63334
6 0.18357 0.63951 -138.78316 -8.31972 0.89782
7 0.23408 1.42265 -138.78411 -8.63945 0.95861
7 0.23408 0.99952 -138.78411 -8.68592 0.76104
8 0.23408 --- -138.78411 --- ---
According to the criterion of the convergence of dimer, the "force" should be lower than the value of "ediffg".
So, didn't I find the saddle piont in this dimer calculations?
If so, how can I improve my calculations to find the real saddle point with dimer method?

p.s. my incar is attached:
Startparameter:
PREC=Normal
NPAR=12
ISTART = 0
ICHARG = 2
INIWAV=1
ISYM=0
ISPIN=1
LWAVE = .FALSE.
Electronic Relaxation:
LREAL=Auto
ENCUT= 350
# ENAUG=400
EDIFF=1.0E-7
NELM=60
IALGO=48
Ionic relaxation:
NSW=400
IBRION =3
POTIM=0
EDIFFG=-0.08

DOS related values :
ISMEAR = 0
SIGMA = 0.1
NEB values:
ICHAIN=2
IOPT = 2

[graeme]

This calculation looks like it's converging to a saddle, but it's not finished yet.
Convergence should speed up now that the rotation is settled down.