Files are estimates of rho using Hudson's maxhap for teosinte, landraces, and maize from Hufford et al. Nat. Gen. 2012 (though these data were not published with the paper). The files are a bit redundant, but each line looks like:
1 100 141 100 14 0.000098
Which is chromosome, window, and number of SNPs repeated twice, followed by the MLE of rho. So the line above would be the 100th 10kb window on chromosome 1 (on reference genome AGPv1), with a rho=0.000098. I would be hesitant to trust the values with low S (definitely <10 and probably <20 <30), as those probably reflect noise more than anything else.