High-resolution mass spectrometry (HRMS) is widely used for the drug screening of biological samples in clinical and forensic laboratories. With the continuous addition of new psychoactive substances (NPS), keeping such methods updated is challenging. HRMS allows for combined targeted and non-targeted screening; first, peaks are identified by software algorithms, and identifications are based on reference standard data. Remaining unknown peaks are attempted identified with in silico and literature data. However, several thousand peaks remain where most are unidentifiable or uninteresting in drug screening. The aims of the study were to apply a combined targeted and non-targeted screening approach to authentic driving-under-the-influence-of-drugs (DUID) samples (n = 44) and further validate the approach using whole-blood samples spiked with eleven low-dose synthetic benzodiazepine analogues (SBA). Analytical data were acquired using ultra-high-performance liquid chromatography coupled with a time-of-flight mass spectrometer (UHPLC-TOF-MS) with data-independent acquisition (DIA). We present a combined targeted and non-targeted screening, where peak deconvolution and filtering reduced the number of peaks to inspect by three orders of magnitude, down to four peaks per DUID sample. The screening allowed for tentative identification of metabolites and drugs not included in the initial screening, and three drugs and thirteen metabolites were tentatively identified in the authentic DUID samples. Running targeted-screening true-positive identifications through the filters retained 73% of identifications. In the non-targeted screening, nine of the spiked SBAs were identified in the concentration range of 0.005-0.1 mg/kg, of which three were tentatively identified at concentrations below those reported in the literature.