Of the wide variety of measures of genetic distance or 
no
single estimator has a superior performance in all situations. This is
due to the complexity of the evolutionary process of the
microsatellites themselves. There are at least two distinct classes of
mutations (single-step and multi-step mutations). The prevalence of one
class of mutation over the other dramatically affects the estimators.
The prevalence of either class of mutation may depend on the structure
of the microsatellite repeat itself. Irregular and composite repeat
structures seem to reduce the amount of single-step mutations, making
the mutational process more similar to the IAM. Futher, simulation
studies find that three to five bp repeats seem to evolve via the SMM,
while 1-2 bp repeats are more similar to the TPM. Although, the
inclusion of results from many loci should increase the power of the
estimators, it is clear that care should be taken not to combine the
results from the above classes of loci without taking these differences
into account.
The comparison of the various estimators from simulated and real data illustrate this point. A general trend emerges, that at short time periods, when the effects of mutation are minimal (i.e. no homoplasy) and the differences in allele frequency are mainly due to drift, the IAM based estimators perform best. However, as the effects of mutation increase, the SMM estimators perform better. Therefore, when the alleles contain no relevant history of mutational events, as might be the case for composite microsatellites, the IAM based estimators perform better than the SMM estimators.
There is much further work that is needed before the mutational processes of microsatellites can be more fully understood and modelled. This review suggests some obvious directions. For example, the existence of a size constraint on microsatellite repeat length and its affects on both the SMM and IAM based estimators needs to be examined. Further, the possible bias of an increase in mutation rate with an increase in repeat length needs to be addressed. Finally, the properties of the private allele method of estimating Nm, under both the SMM and TPM, should be examined through simulation to assess the usefulness and limitations of the estimator.