SBS signatures choice

[N-TI]

Hi,

I'm analyzing SBS signatures one tumor type (extract known mutational signatures). First, I did the fitting to search signatures with real contribution (I have extracted all the cosmic signatures in input_signatures parameter ).
However I found ambiguous results. For example, SBS3 doesn't appear as a signature with real contribution (very few samples present the SBS3 with low contribution) . However, when I choose some known signatures (8 or 10) and I include SBS3, I've got another totally different result: I have found that SBS3 has appeared in the most of samples with very high proportions.

As I'm beginner in signature analyses, my question is why I had two different results for the same samples just because I have changed the number of signatures to extract and how can I fix this problem?

Thank you!