GCTA conditional analysis - too many signals?

Hello,

I am using GCTA to identify independent signals within a 2MB region surrounding my top SNPs from a GWAS.  The command I am using is:

gcta64 --bfile signals/rsID1 --cojo-file signals/results.ma --cojo-slct --cojo-actual-geno --cojo-p 1e-05 --out conditional/rsID1 --thread-num 10

For most SNPs I get the original SNP back with no other independent signals.  However, for a subset of (usually low-frequency, MAF ~ 0.01) SNPs the list returned contains >100 (very rare) variants and does not include the original SNP.  The very rare variants were not significant in the original GWAS.  Does this imply the original SNP was only acting as proxy for a large number of rare signals with a large combined effect on phenotype?  Or am I doing something wrong?  Should I be setting an MAF threshold for example?

Thanks

Kath