Hey Aaron, this could be an artifact of PCA.  The linear nature of PCA sometimes causes what some refer to as the "horseshoe effect" on PCs 1 and 2, rendering them not completely independent...  I don't know anything more about it mathematically, but using that term (here or in Google) may help you find the real answer for your situation. 

Here is one such google hit that discusses the issue: http://phylonetworks.blogspot.se/2012/12/distortions-and-artifacts-in-pca.html

Hi Aaron,

If I understand well your plot the points are individuals. With my very little knowledge of PCA, I would say:

1_that you have several outliers individuals that shows rare snp alleles.

2_as these individuals and/or rare alleles are at a very low frequency compared to the rest of the population (large number of individuals) this outweighs the first 2 axis of the analysis and skewed their distribution.

3_these alleles and or individuals could be problematic (errors) or very interesting (structure, selection ... who knows).

4_Try to identify both individuals and alleles is what I would do in that case (might be very easy based on their individual eigenvalues for the first 2 axis).

Hope this helps

By the way I am not statistician, my answer may be very false. :/

See you soon,

Jordi

Aaron

Here is a classical paper that discusses many of the problems associated with multivariate analyses including PCA. It's fairly old now but it addresses the problem you are encountering with your data and obviously you can cite it if need be.

Cheers.