The application of the template-based correction for partial volume effect (PVE) may be challenging in clinical situations where a non-homogeneous region of interest (ROI) is surrounded by a non-homogeneous background. In this paper, we examine the performance of our iterative PVE correction (itPVEC) under such circumstances. We simulated a number of oncology SPECT scans where four tumour-like targets were adjacent to the tissues with different activity concentrations. We considered three types of activity distributions inside ROls (1) the cone-shaped non-uniformity where an inner elliptical core had higher concentration than the outer shell; (2) the concave-shaped non-uniformity where an inner elliptical core had lower concentration than the outer shell; (3) locally uniform distribution where the target was composed of two parts with different but uniform activity concentrations. We simulated a typical oncology SPECT scan with an acquisition matrix of 128x128x128 and pixel size of 4.4mm. We assumed that the boundaries of our targets were available from the high-resolution anatomical image with a matrix 512x5I2x5I2 and voxel size of l.Imm. The template based partial volume effect correction demonstrates the best performance when the true activity is uniformly distributed throughout ROI. However, this approach is worth to apply even in situations with non-uniform activity distributions (inhomogeneous tumours). Based on our analysis of the PVEC applied to three models with four different tumour-like homogeneous and inhomogeneous targets, we conclude that template-based method improves (i) total activity (errors did not exceed 6%) and (ii) distribution near tumour boundaries.Especially, it can substantially improve imaging of non compact tumours.