Using resting-state fMRI is a challenge in patients with large tumors owing to mass effect and potentially altered representation of function, i.e., remapping. The best approach to atlas registration in tumor patients remains uncertain. In this work, we investigated the effects of alternative approaches to atlas registration on resting-state fMRI functional connectivity in glioblastoma (GBM) patients and studied the impact of GBM on functional brain organization. We analyzed data representing 59 GBM patients and 161 age-matched healthy subjects and systemically investigated affine vs. non-linear atlas registration and two associated masking options. Outcomes were assessed both in terms of structural data match to an atlas template as well as functional connectivity (FC) match to a reference dataset. Next, we evaluated the impact of GBMs on the organization of brain networks accounting for the hierarchical organization of resting state networks. Our results demonstrate that non-linear atlas registration with tumor masking optimizes both structural normalization and FC mapping. The aberrancy map results, obtained with fine parcellations, are consistent with prior findings showing that FC abnormalities induced by focal lesions are widely distributed. However, the results obtained with coarse parcellations indicate that functional architecture is grossly normal in tumor patients.