how to select the best combination of gap_open and gap_extention in pairwise sequence alignment?

Question

[imlituan]

I recently use clustal omega to make a multiple alignment and get the pairwise identity matrix. Then I use the needle to align two sequences from this msa. I find that, the identity from needle alignment is lower than that from msa. I think it may be the result of bad parameters of needle. Then i write a script using pairwise2 from Biopython, and try lots of combinations of gap_open and gap_extend. From each combination, i get and score and identity, then make two surface plots. The results show that score and identity is not always agree with each other. Another point is that, when gap_open and gap_extend are both very lower, the score is higher. So I wonder, should i pick the combination with the best score or best identity.