Separation techniques hyphenated to high-resolution mass spectrometry are essential in untargeted metabolomic analyses. Due to the complexity and size of the resulting data, analysts rely on computer-assisted tools to mine for features that may represent a chromatographic signal. However, this step remains problematic, and a high number of false positives are often obtained. This work reports a novel approach where each step is carefully controlled to decrease the likelihood of errors. Datasets are first corrected for baseline drift and background noise before the MS scans are converted from profile to centroid. A new alignment strategy that includes purity control is introduced, and features are quantified using the original data with scans recorded as profile, not the extracted features. All the algorithms used in this work are part of the Finnee Matlab toolbox that is freely available. The approach was validated using metabolites in exhaled breath condensates to differentiate individuals diagnosed with asthma from patients with chronic obstructive pulmonary disease. With this new pipeline, twice as many markers were found with Finnee in comparison to XCMS-online, and nearly 50% more than with MS-Dial, two of the most popular freeware for untargeted metabolomics analysis.