We investigated the quantitative accuracy of SPECT/CT imaging studies as would be performed before and after targeted radionuclide therapy (TRT) using phantom experiments with (i) (99m)Tc, (ii) ¹⁷⁷Lu and (iii) ⁹⁰Y/¹⁷⁷Lu. While the experiment with (99m)Tc imitated a diagnostic scan, the experiments with ¹⁷⁷Lu and ⁹⁰Y/¹⁷⁷Lu modeled post-therapy acquisitions. At the next stage, we reconstructed images from pre- and post-therapy patient studies. The data were first reconstructed using two methods with limited corrections for the physics effects. Then, to generate quantitatively accurate absolute activity distributions, we applied a hybrid (model-based and window-based) reconstruction strategy where some of the physics effects were accurately modeled while corrections for other effects were empirical and based on information obtained from the projection data. The accuracies of absolute activity recovered by the hybrid method from the six phantom experiments were very similar to each other and acceptable for potential use in TRT. When measured in identical regions of interest, the (99m)Tc ⁹⁰activity was reconstructed with errors ranging between -3.3% and 2.9%, while the ¹⁷⁷Lu activity was reconstructed from experiments with ¹⁷⁷Lu and Y/¹⁷⁷Lu with errors ranging between -1.6% and 1.6%. The reconstruction algorithms with limited corrections led to larger and case-specific errors as might have been expected. From a clinical prospective, our results showed that physics-based reconstructions improved resolution of images corresponding to both diagnostic scans with (99m)Tc and post-therapy scans with ¹⁷⁷Lu. Our analysis of patient study demonstrated that lack of corrections led to overestimation of activities in organs and tumor by 29-39% for the diagnostic scan with (99m)Tc and by 105-218% for post-therapy scan with ¹⁷⁷Lu.