I did a TS calculation for the CH4->CH3+H step on Ni(111) surface. The method chosen was complete LST/QST with 20 (maximum number of steps is 20) QST steps, fine quality and RMS convergence as 0.05 eV/A. After the 20 steps over I got the optimized TS structure where some but not all parameter matched with the previous work in the literature. However the RMS convergence achieved after 20 QST steps was only 0.44667 eV/A which is far away from the fine settings. I have also added the text below taken from the .castep file.

QST cycle limit reached
 QST cycle:                                      20
 Maximum number of QST cycles:                   20
 Transition state search failed
 Please check the endpoints
 Cumulative number of energy/gradient calls:    505

I therefore request you to help me achieving the convergence of fine settings (0.05 eV/A). What modifications do I need to do? And also let me know if there is any way to restart the calculation after such results because ultimate task is to achieve desired convergence which I am not able to get.
Thanking you in advance,