long haplotype as population structure

[amsamani]

Hi,

I am looking at a very closely related population and I realized that p-value for a really important region is very reduced in univariate linear mixed model analysis with relatedness matrix. I looked at the region and I realized that the haplotype is at least 4Mb long in this specific population. would that be a reason for the fact that I am not seeing an elevated peak at that region with GWAS?
In general, how long should the haplotype be so that GEMMA doesn't mark the region as population structure.

Thanks.