Well first i guess you have to check for potential technical problems, e.g. microsatellite pattern and stutter bands might lead to incorrect reading of the alleles, null alleles and so on. Once you have checked for this, you can find or try to find biological explanation... So the first step is to check the loci and you can use microchecker for that.
It would also be interesting to know the levels of FIS you are getting for these microsatellite loci. Now knowing that some Genomics Islands exist and that some of your loci might be on this part of the genome, I would definitively assess if they are under selection and what kind of selection (you can use LOSITAN or BAYESCAN).
But of course everything depends first on the technical aspect of it... I would also suggest you to do analyses with and without these loci to see if the genteic pattern you obtained is different.
I agree with checking the technical problems first. Given the organism and it's likely population structure in nature, I would guess a population / biological explanation is very likely. Remember that we simply use HWE as the null model and deviations from HWE tend to be really interesting (selection, drift, non-random mating...) in the modern landscape.