Dear Thripuranthaka,

please have a look at the atomic numbers of the constituents of your sample:

Cu (Z=29) , Sn (Z=50) and  S( Z=16).

The fluorescence line energies are around 8 to 9keV for Cu (K shell excitation),

around 29 keV for Sn (K Shell excitation) and 3 to 4 keV for L shell excitation;

for S the fluorescence line energy is around 2,3 to 2,4keV.

So the fluorescence of S is not really able to escape from a matrix set up of quite heavy elements (and therefore stongly absorbing) such as Cu and Sn compared to S.

Although for the Sn L-shell based fluorescence the situation is a little bit better, but not really good in comparison with the Cu fluorescence case.

Due to these reasons it is no wonder for me  that a standard software cannot really evaluate quantitatively the relative atomic concentrations (At%) of this sample.

These links will be useful to you.

http://emu.uct.ac.za/training/sem-school/precision-and-accuracy-in-edx-analysis/

http://cfamm.ucr.edu/documents/eds-intro.pdf

Best regards.

Dear Dr. Gerhard sir, 

Thank you very much for your kind response. 

In my question I forgot to mention that, I did EDAX of the same sample in another SEM in which the At% of the Cu, Sn and S followed quite closely. This was done at 15 kV. and in another case in which the At% was not followed also we did both at 30 kV, 15 kV and 10 kV but still we observed the issue. 

Dear Professor Yuri Mirgorod, 

Thank you very much for providing the useful links. 

I you want quantitative result, you have to use standards. Standardless analysis is prone to errors, especially (as Gerhard Martens has noted) when difference in atomic numbers of constituents is quite big.