Closed:SNPeff: can I just use hg19?

Question

[Marvin]

If I haved used hg38 to call variants and then use SNPeff in combination hg19 to annotate that file, what will happen?

I would expect the output to be totally wrong. Because my positions in the VCF refer to hg38. It could well be that my SNP, which is at position 5000 in hg38 is at position 4000 in hg19 right? So SNPeff would check position 5000 in hg19 instead of position 4000. Am I right with my thinking?

Then why does the SNPeff documentation recommend to use the latest version of hg19? How can you make such a general statement? I think the reference HAS TO BE the same that was used to call the variants or am I wrong here? Because using another reference would result in column 2 of the VCF indicating wrong genomic positions, right??