Substantial insertions and deletions often happen by mechanisms that are pretty well understood - non-allelic homologous recombination between repeated sequences, insertion of transposable elements, and others. It’s true that there’s no way to know for certain how many mutations occurred to produce the sequence differences between species we see today. Generally parsimony is the rule for analysis. Ideally you determine the ancestral sequence by looking at chimp, gorilla, orang, etc and taking the most parsimonious mechanism to get to the pattern seen today. I’ve seen articles by anti-evolution guys crowing about all the positions where human and gorilla have one nt and chimp another. The rather obvious conclusion is that most of these are mutations that are chimp-specific.
Different kinds of mutations have rates that vary a lot. Generally repeat copy number changes occur at much higher rates (events per generation) than point mutations. Changes in repeat copy number in microsatellites (short tandem repeats,) the stock in trade of genetic genealogy, are an example. A few larger repeats change copy number at such high rates that they will often distinguish monozygotic twins.
Small insertion/deletion events not involving repeats happen at ~10 times lower rate than point mutations. The larger an indel size you look at, the less frequent they are, unless they are repeat based. If you really wanted to be painstaking, you would look at the most likely event type to produce each sequence difference, take its estimated rate and add them up for all the events to estimate age of the common ancestor. But SNP rates are much better studied than the others, and there are far more SNPs, so that is the most straightforward way to do the calculation.
I just cooked up a way to calculate genetic distances for genealogy purposes that weights the copy number difference at each STR locus by the inverse of its measured mutation rate. When someone has changed copy number at several slow mutating STRs it makes a big difference in the results. STRs do weird things though - the only reason for messing with them is that that’s what most people have had measured. SNPs (single nucleotide changes) are much better behaved and more useful for estimating the time of the the most recent common patrilineal ancestor of living people.