Transient decreases in the proportion of individuals newly infected with an HIV-resistant virus (primary resistance) are documented for several cities of North America, including San Francisco. Using a staged SI deterministic model, we identified three potential causes consistent with the history of the epidemic: (1) increase in risky behaviour, (2) reduction in the proportion of HIV-acutely infected individuals undergoing treatment, and (3) replacement of mono- and dual-drug therapies with triple-drug therapies. Although observed patterns resemble scenario 1 most closely, these explanations are not mutually exclusive and may have contributed synergistically to the decline. Under scenario 1 the counterintuitive situation arises where, although the proportion of primary resistance cases decreases transiently, the epidemic worsens because the actual numbers of infected individuals and of drug resistance carriers increases. Our results call for improved efforts to control the epidemic in developed nations, and highlight the usefulness of drug resistant strains as epidemiological markers.