No I haven't found any direct solution to that problem with PGD Spider. Instead of trying to fix it, since I didn't have any answer, I created my own custom Python scripts to parse the VCF file, produce a genotype matrix file and then parse this genotype matrix file to create the input file for Bayescan.
The only problem is that my scripts are extremely custom, i.e adapted to my file names and population names. If you are completely stuck with this, you can quickly tell me what kind of data you have, how many populations and maybe give me the header of your VCF file (including the columns that contain the names of your SAM files, i.e all the lines that start with "##" and the line that starts with "#CHROM").