Matt,

I suspect these reflect different sequence settings on teh Philips console. I assume these are all true gradient echo fieldmaps (rather than PEpolar). I do not have access for Philips hardware, but on Siemens it is important to set the reconstruction to save both Magnitude and Phase:

  https://github.com/neurolabusc/dcm_qa_fm...

While I am less familiar with the details, you can also set what data is exported, as seen in the ADNI data described on this web page. I also provide details on how to proceed with visual inspection:

  https://crnl.readthedocs.io/fieldmaps/in...

Base on your data, you may want to preprocess it:
  https://web.mit.edu/fsl_v5.0.10/fsl/doc/...

Unfortunately, Philips DICOM images are impoverished with crucial details for understanding the sequence. This is a limitation of the source images, not dcm2niix. Therefore, you will need to add details to provide BIDS compliant datasets. I describe this here:

 https://github.com/rordenlab/dcm2niix/tree/master/Philips#missing-information

My advice is to work with the Philips Clinical Scientist who is affiliated with this magnet.