Dear Santanu Pathak,

          you are trying to refine the atomic positions of a lighter atom. First thing come to my mind is that how light is your atom? . If it is a low z element, there will not see much difference in the agreement indices, which you get after the refinement. The reason is obvious,  the contribution from a lighter atom is less/null in your xrd pattern. The problem of toggling of atomic positions could be arising due to the falling of parameter value in the local minimas. If that is the case, one can toggle between these two local minimas by applying a small  but a definite shift, which the programm calculates. The damping option provided in the GSAS programm allows you to control the shift applied to the parameters after each refinement cycle. you may use this option to stabilize the refinement of atom positions. In this case, you are partially shifting the parameter to get a stable refinement. If you are using the damping parameter, D, in your refinement, the shift applied to the parameter will be

   = original shift*(1-D/10);

if you give D=6, 40% of your original shift will add to the parameter.

After all, keep in mind that, the atom positions of llow z atoms refined from x-rd is less reliable.

Dear Irshad K A, Thank u for your suggestion. In my refinement I am meaning oxygen (O) as lighter atom. And after toggling in few cycles both the set of values are being stabilized to 2 different sets of co-ordinates.

Dear

Santanu

I am completely agree with Irshad K A .Fluctuation of atomic position of lighter atoms like oxygen is frequent case during Rietveld Refinement. You can also control it by choosing suitable value of 'B' factor. Every time you can measure bond lengths. Is it compatible with your other results like magnetization or properties which are strongly related with bond lengths?. Don't care about chi^2.