Although existing theory suggests that limited evolvability drives self-compatible plants to an evolutionary “dead end,” there are in reality many self-compatible plants. One reason for this discrepancy may be that the advantage and disadvantage of self-compatibility are different between evolutionary and population-dynamical timescales. Here, we constructed a spatially explicit computer simulation model to describe the spatial dynamics of multi-species metapopulations with phylogenetic dynamics of speciation, extinction, and transition from self-incompatibility to self-compatibility. The model assumed spatial gradients and temporal fluctuations of the environment of habitat patches, and that local populations can adapt to their local environments. Self-compatible populations were assumed to have limited standing genetic variation, and thus lower evolvability, than self-incompatible populations. On the other hand, because of mating assurance, self-compatible populations were assumed to have a higher potential to colonize in empty patches. Simulations found that when the environmental gradient was shallow, self-compatible species dominated the whole landscape; however, when the environmental gradient was steep, whether self-compatible or self-incompatible species dominated depended on the colonizing potential of self-incompatible populations. These results showed that self-compatible species continue to exist because of their advantage on a population-dynamical timescale (i.e., mating assurance) exceeds their disadvantage on an evolutionary timescale (limited evolvability).