Your "small bump" as you call it is not a big deal in a GWAS context, since it grossly corresponds to a unadjusted p-values = 0.01 and won't be considered after multiplicity correction. Another thing you can check is your p-value histogram. However if you want to remove this little jump above the straight line, depending on the model considered you can for example detect if there are additional sources of bias incorporating more covariables. Population structures in GWAS is often a big source of problems. You can edit your post with additional details on the model used, histogramm, etc for a more comprehensive answer.