Hi Naveen,
In fact, the Taylor expansion of the cos function up two second order doesn't contain more than the first nearest neighbor hopping, so the result doesn't correspond to the tight binding model you are comparing with. You should expand the function up to forth order, then you will be able to see second nearest neighbors from the kwant discretization.

Regards,
Ousmane