Hi sangita, While modelling a protein we generally look at the sequence identity. E value doesn't give you clear idea about the sequence, specially in case of shorter templates.
The threshold for the sequence identity is 50% more or less. Anything below that will give you a bad model.