For the moment, ignore the chromatogram traces in Xcalibur please. What are the actual spectra? Some are lock-mass corrected and some are not? Is it like every scan is collected once uncorrected and then stored again as corrected? If so, do the scans have 2 different retention times?
We'll probably need to see the data file (can you upload it?), but if the scans have different RTs, then I suspect Skyline will be using both uncorrected and corrected scans. There's nothing in ProteoWizard or Skyline to check for this case. And it seems quite an odd one. If each scan is duplicated, your files will be twice as large but without any real benefit (unless of course you are trying to evaluate the benefit you get from lock mass correction) :)