The conventional way to analyze methylation data is to look for genomic regions that have consistent changes in methylation status, as biologically it's expected that regional changes in methylation are more likely to have an effect than changes for a single CpG. In addition, methylation arrays are based on differential hybridization of probes, and as long experience has taught us, the hybridization of short sequences of DNA are affected by far more than just complementarity. In other words, the signal from a single probe may or may not completely reflect underlying biology.
You might consider using any of the packages intended to infer region-based differential binding rather than relying on single CpGs. The next step is usually identifying genes that may be close enough to the regions of differential methylation to be affected, using things like CHiPpeakAnno or CHiPseeker.
