Help! I am doing some regular genotype population stratification using smartpca from EIGENSTRAT 5.01. The plot of PC2 ~ PC1 looks like the picture. The color ones are hapmap (blue/lightblue: euro; greenish: asian; redish: blacks; purple: mexican). The grey ones are my samples. There is a shift of coordinates, which indicates something wrong. Does anyone have any idea about the problem?

Details of what I did: I got my sample genotypes, merge with hapmap ones. There are some snp with strand flips, and I have tried both flip or just filter out. Also I have tried to whether remove long-range LD regions; whether to prune local LD SNPs, etc. No matter how I do it, the figure looks like this.