Magnetic susceptibility artifacts, including both image distortions and signal losses, exist near air/tissue interfaces in the ventral brain in standard blood oxygenation level dependent (BOLD) functional magnetic resonance imaging (fMRI). Although several acquisition-based approaches exist to address the signal losses, they require increased acquisition time or patient customization. In this work, we propose a statistical estimation model that includes the effects of magnetic field gradients (both within-plane and through-plane gradients) and uses an iterative reconstruction algorithm to reconstruct images corrected for both magnetic field distortion and signal losses. Besides, we combine our reconstruction approach with a recently proposed MRI sequence with Z-shimming gradient between the spiral-in and spiral-out acquisition to enhance the compensation for signal losses. Therefore, we extend our forward MR signal model to include the physics of Susceptibility-induced magnetic Field (SF), Susceptibility-induced magnetic Field Gradients (SFG), and the application of the data acquisition technique with Z-shimming Gradients (ZShG). The results show that not only signal distortions but also significant signal losses can be compensated by considering both the modeling of field-inhomogeneity effects along with the acquisition using Z-shimming.