This study is focused on measuring the densities of the excited molecular oxygen species, O2(a1∆g) and O2(b1Σ+g), produced in a COST atmospheric pressure plasma jet using a helium-oxygen mixture. Knowledge of the ozone density is critical for measurements because of its high quenching rate of these species. Additionally O2(a1∆g) is difficult to measure, due to its low emission intensity and sensitivity to background interference in the plasma region. Therefore a flow cell was used to enhance signal detection in the effluent region. To validate the measurements and improve understanding of reaction mechanisms, results were compared with two simulation models: a pseudo-1D plug flow simulation and a 2D fluid simulation. The plug flow simulation provided an effective means for estimating species densities, with a fast computation time. The 2D simulation offered a more realistic description of the flow dynamics, which proved critical to correctly describe the experimental trends. However, it requires long computation times to reach an equilibrium state in the flow cell. Otherwise, it leads to discrepancies to the experimental data. Further discrepancies arose, from an overestimation of the ozone density from the models, as validated from the O2(b1Σ+g) density measurements. Optimizing the reaction rate coefficients for the effluent region might improve the agreement with the experimental results. Despite these limitations both simulations aligned reasonably well with experimental data, showcasing the well validated plasma chemistry of the models, even for complicated effluent geometries.