Clinical markers in the peripheral blood guide the treatment of human immunodeficiency virus type 1 (HIV-1). Likewise, many of the theoretical models developed to simulate infection only incorporate variables in the blood. To test the suitability of blood-only models, three distinct models of HIV infection kinetics are compared: "full model" including latently and actively infected cells and virus in the peripheral blood and lymphoid tissue (LT); "reduced model", including peripheral blood and LT without latent cells; and "blood model" including only actively infected cells and virus in the peripheral blood. Using the same parameter values for all three, qualitative differences are demonstrated between the blood model and its more inclusive counterparts. Additionally, optimization studies show that the reduced and blood models generate progressively lower optimal treatment levels relative to the full model when constant-level treatment is considered. These findings indicate that including the lymphoid tissue and latently infected cells into kinetic models may lead to differing conclusions with regard to optimal treatment and could be useful in guiding therapy even when plasma viral levels are below detectable limits.