This study systematically evaluates 768 data-processing pipelines for network reconstruction from resting-state functional MRI, focusing on brain parcellation, connectivity definition, and global signal regression. The evaluation aims to identify pipelines that minimize motion confounds and spurious test-retest discrepancies while being sensitive to inter-subject differences and experimental effects. The results reveal significant variability in pipeline performance, with most pipelines failing at least one criterion. However, a subset of optimal pipelines consistently satisfies all criteria across different datasets, spanning various time intervals. The study introduces a comprehensive framework for evaluating network construction pipelines, including an information-theoretic measure called "Portrait Divergence" (PDiv) to assess network topology. The optimal pipelines are characterized by their ability to minimize spurious differences, detect true experimental differences, and avoid motion-induced confounds. The findings highlight the importance of considering entire pipelines rather than individual steps, as the performance of a pipeline is influenced by the combination of all steps. The study provides a set of benchmarks for trustworthy functional connectomics, emphasizing the need for careful selection of data-processing pipelines to ensure reliable and valid results.This study systematically evaluates 768 data-processing pipelines for network reconstruction from resting-state functional MRI, focusing on brain parcellation, connectivity definition, and global signal regression. The evaluation aims to identify pipelines that minimize motion confounds and spurious test-retest discrepancies while being sensitive to inter-subject differences and experimental effects. The results reveal significant variability in pipeline performance, with most pipelines failing at least one criterion. However, a subset of optimal pipelines consistently satisfies all criteria across different datasets, spanning various time intervals. The study introduces a comprehensive framework for evaluating network construction pipelines, including an information-theoretic measure called "Portrait Divergence" (PDiv) to assess network topology. The optimal pipelines are characterized by their ability to minimize spurious differences, detect true experimental differences, and avoid motion-induced confounds. The findings highlight the importance of considering entire pipelines rather than individual steps, as the performance of a pipeline is influenced by the combination of all steps. The study provides a set of benchmarks for trustworthy functional connectomics, emphasizing the need for careful selection of data-processing pipelines to ensure reliable and valid results.