The cutoff range varies. In some papers i have seen the cutoff to be 60% identity.

See the referrence.

"Subtractive genomics approach to identify putative drug targets and identification of drug-like molecules for beta subunit of DNA polymerase III in Streptococcus species."

Georrge JJ, Umrania VV.

Appl Biochem Biotechnol. 2012 Jul;167(5):1377-95.

Hope this helps !!!

You cannot think of "e" in terms of some universal cutoff value to be applied in all cases. "E" is simple the number of hits one can expect, given your query sequence, searched against a specific database of size N. E takes into account not only the database size, but the length of the query sequence - very short query sequences will have an inherently higher e value than longer query sequences against the same database, so your assignment of hit significance will vary in such cases.

Steve Palumbi has some good guides on using and interpreting BLAST results:

http://sfg.stanford.edu/BLAST.html

You should also have a look at the BLAST documentation at NCBI, and specifically

http://www.ncbi.nlm.nih.gov/BLAST/tutorial/Altschul-1.html#head2 and the references therein.

Basically, you need to understand the statistics of BLAST scoring, and then make informed judgements for your particular data and searches.

Thank you Dipon and Michael. Michael your answer uas really satisfactory and the links were really informative to update my knowledge.