I'm referring to the text described here:
These tools [NCBI remap, CrossMap] operate only on the sites present in an input VCF, and return the representation of those sites in a new genome assembly. This does not capture all variation, however. Consider an individual who has sequence reads that indicate they match the GRCh37 reference genome assembly at position GRCh37.chr1:169,519,049 (i.e. the individual's genotype is T/T). Because the individual is homozygous reference at that site, there will be no variation present in their VCF file created on GRCh37. However, the analogous position on the updated GRCh38 reference genome assembly, position GRCh38.chr1:169,549,811, has the reference base C. Consequently, if the individual's read data were analyzed on GRCh38, they would be identified as homozygous for a C->T SNP. Because this site is not present in the input GRCh37 VCF, it is never added when creating a GRCh38 VCF by these other tools.
How important is this? Do you think it is relevant to any types of analysis?