different predictions for the same poses when different number of frames are used

[shalinia93]

I am using the code to predict results for an intron receptor and a small molecule docked to it.
command: ~/RNAPosers/src/run.sh ./receptor.mol2 ./poses.sd ./out.txt 50
Here, the default values for RMSD and Eta are taken by the code (2 and 248), and 50 frames are used for calculation- below is the out.txt snippet:
----output-----
0.000000 0.567000 0.433000
0.000000 0.588000 0.412000
0.000000 0.582000 0.418000
0.000000 0.581000 0.419000
0.000000 0.595000 0.405000
0.000000 0.591000 0.409000
0.000000 0.570000 0.430000
0.000000 0.582000 0.418000
0.000000 0.581000 0.419000
0.000000 0.568000 0.432000
0.000000 0.794000 0.206000
........
However, when I use 18 instead of 50, the predictions for the same pose vary:
~/RNAPosers/src/run.sh ./receptor.mol2 ./poses.sd ./out.txt 18
----output-------
0.000000 0.641000 0.359000
1.000000 0.011000 0.989000
1.000000 0.010000 0.990000
1.000000 0.011000 0.989000
1.000000 0.067000 0.933000
1.000000 0.062000 0.938000
1.000000 0.012000 0.988000
1.000000 0.017000 0.983000
1.000000 0.009000 0.991000
1.000000 0.061000 0.939000
0.000000 0.785000 0.215000
.......

The same happens when I use 100 poses where everything is consistently 0. I was wondering if any parameters need to be tuned to obtain consistent results.

Thanks,

Shalini