Imatinib and other tyrosine kinase inhibitors (TKI) have improved treatment of chronic myelogenous leukemia (CML); however, most patients are not cured. Deeper mechanistic understanding may improve TKI combination therapies to better control the residual leukemic cell population. In analyzing our patients' data, we found that many patients who otherwise responded well to imatinib therapy still showed variations in their BCR-ABL transcripts. To investigate this phenomenon, we applied a mathematical model that integrates CML and an autologous immune response to the patients' data. We define an immune window or a range of leukemic loads for which the autologous immune system induces an improved response. Our modeling results suggest that, at diagnosis, a patient's leukemic load is able to partially or fully suppress the autologous immune response developed in a majority of patients, toward the CML clone(s). Imatinib therapy drives the leukemic population into the "immune window," allowing the patient's autologous immune cells to expand and eventually mount an efficient recognition of the residual leukemic burden. This response drives the leukemic load below this immune window, allowing the leukemic population to partially recover until another weaker immune response is initiated. Thus, the autologous immune response may explain the oscillations in BCR-ABL transcripts regularly observed in patients on imatinib.