In constant population size models, we observed only a slight reduction in ARG inference accuracy in terms of estimates of coalescence times or recombination breakpoint counts when using computationally phased haplotypes. The effect also reduced as the sample size increased. 4/7
