Background: Liquid biopsy matrices-including plasma, serum, saliva, and urine-are increasingly used for non-invasive cancer detection, companion diagnostics, and recurrence monitoring. However, conventional column-based purification and double-stranded DNA library preparation fail to capture the full cfDNA landscape, particularly ultra-short fragments and single-stranded DNA (ssDNA). We evaluated a novel nucleic acid purification method paired with an ssDNA-compatible library preparation workflow to enable more comprehensive cfDNA profiling, serum and saliva samples as model systems. Results: We compared several commercially available bead-based cfDNA extraction methods. The MAGicBead cfDNA Isolation Kit, which incorporates a novel nucleic-acid-binding surface, yielded the highest cfDNA recovery. Recovered cfDNA was processed using a library preparation workflow derived from the original Splinted Ligation Adapter Tagging (SPLAT) chemistry, enabling capture of both dsDNA and ssDNA. A distinct population of ultra-short DNA fragments (35-75 bp) was observed in the mapped reads and was detected only with MAGicBead purification, while the dominant nucleosome-associated peak was retained.To determine whether these ultra-short fragments were single-stranded, isolated cfDNA was treated with a single-strand-specific nuclease prior to library preparation. Following digestion, the short-fragment signal was no longer detectable, confirming their ssDNA nature and underscoring the need for an ssDNA-compatible SPLAT-derived workflow.We next applied the optimized workflow to plasma from lung cancer patients and observed differences in the proportion of ultra-short cfDNA fragments relative to healthy controls, suggesting potential fragmentomic utility.Because saliva is an emerging but technically challenging liquid biopsy matrix due to mixed human-microbial content and heterogeneous fragmentation, we further evaluated the workflow in saliva. The ssDNA fragment population was again recovered, and sequencing simultaneously revealed microbial signatures, demonstrating that the workflow captures both host and microbial DNA in a single run and supports a multi-omic approach that may enhance diagnostic performance. Conclusions: The MAGicBead cfDNA Isolation Kit, combined with SPLAT DNA library preparation, enables robust recovery of ultra-short cfDNA fragments that are routinely lost with conventional workflows. This approach supports comprehensive cfDNA profiling across and below the nucleosome-size range. Our pilot studies reveal previously unrecognized cfDNA features and point toward more sensitive strategies for cancer diagnostics and biomarker discovery.