Ks saturation

Dear all,I am trying to calculate dn/ds values of  all orthologous gene pair between a pair of genome
using pairwsie_kaks.pl script within bioperl which evokes the codeml program in runmode -2. When I am
analyzing the results, some of the genes have anomalously high dS or Ks values and some of them even
reaching more than 100 as a result of which the average value of Ks for the whole genome  shots up.
These genes are orthologous genes and even share more than 50% sequence identity. Should I consider these
genes for the anlysis or left them out. If I left them out, then upto what cutoff value of Ks should I consider
for analysis. In some papers, I have found that they have considered Ks values as high as 5.6. Is there a way
for determining the cutoff value for Ks?
Subarna