A modern result converts to an RFLP band only if you recompute the fragment the old probe would have seen. The sequence or SNP file does not already contain band sizes.
Autosomal RFLP (the kinship bands). Pick the historical locus, enzyme, and probe coordinates. From a VCF or aligned reads, take every variant inside that interval that changes a cut site or the length between cuts: a SNP that creates or destroys the enzyme motif, or an indel or VNTR copy-number change. Digest the person’s sequence in silico, keep only fragments the probe overlaps, and report those lengths in kb. A homozygote is one band; a heterozygote is two. That kb value is the comparable band. Repeat-count alleles from old forensic VNTRs convert the same way: flanking distance plus (repeat length × copy number).
mtDNA haplogroup (the +/− site list). A modern haplogroup call already implies the classic sites. Map each diagnostic SNP back to the enzyme motif it creates or destroys and score it as the old papers did: +HpaI 3592, −AluI 10397, 9 bp del 8281–8289. The string of pluses and minuses is the RFLP haplotype; the haplogroup name is just the label for that motif. Y-chromosome haplogroups work the same if the old marker was a restriction site.
What does not convert cleanly is a chip genotype outside those cut sites, or an STR profile from a locus that was never an RFLP. Those have no band to recover unless you sequence the actual interval and run the digest above.