The field of connectomics has introduced new computational tools to the system neuroscience domain, facilitating novel investigations into the mechanisms of the human brain. While methodological advancements in connectomics have found numerous applications with exciting interpretations, a systematic analysis on the reliability and reproducibility of the results and inferences drawn from the findings is still missing. We aim to fill this gap by studying the reproducibility of connectomics findings on diffusion data that has been collected at different sites with varying acquisition protocols. Structural network properties at local, meso, and global scales were investigated to determine what is different and what is preserved in the presence of site related variation. Our comprehensive investigation revealed that despite significant differences between local and global properties, the topological features were mostly preserved.