Unexpected large number of CpG sites

[Larrycpan]

I have a methylation data sequenced with a methylome panel. However, when I look into the coverage/bedgraph file generated by bismark_methylation_extractor, the number of CpG sites exceed than the panel contains, even if I have subset the BAM file based on the panel bed file. Does this make sense? Or should I perform additional filtering (e.g. remove either methylated or unmethylated cytosines less than 30 using the bedgraph file)? Look forward your reply.