Direct-infusion shotgun proteome analysis (DI-SPA) increases throughput by removing liquid chromatography, but the resulting DIA spectra are highly multiplexed and difficult to interpret. This dataset was generated to evaluate ProjDIA, a computational workflow for direct-infusion DIA proteomics. ProjDIA uses fragment-bin projection scoring, confidence-guided two-pass mass-error recalibration, monotone q-value reporting, and semi-supervised rescoring to improve peptide and protein identification confidence. The dataset includes benchmarking experiments across MS2 resolution and sample load, replicate-series analyses, species-entrapment controls for external error assessment, and a three-species mixture for label-free quantification evaluation. The submission contains the raw mass spectrometry files together with processed search and result files used to support the analyses reported in the study.