First of all, I would suggest you make sure you are using the latest version of MRIcroGL and dcm2niix (20190902), as the ability to infer GE slice Timing has improved. Details regarding how dcm2niix estimates slice times for GE data are provided here. As noted there, GE does not provide a universal and unambiguous way to determine slice times. Therefore, we need to guess based on proprietary data that is poorly documented. While the formula seems to work, it may be incorrect for situations such as multi-band sequences. 

In brief, this reflects a limitation of the GE DICOM images, not of dcm2niix. We do not have any GE hardware at my center. However, I would encourage users who use GE equipment for research to lobby their GE Research Collaboration Managers to use public GE DICOM tags to explicitly report features that are important for scientists (slice timing, readout time, etc).