<p>Laser-induced breakdown spectroscopy (LIBS) is gaining prominence in analytical chemistry for direct elemental analysis in solid samples, although its sensitivity remains limited, typically ranging from 1000 mg kg⁻¹ to 100%. Error structure utilization can improve accuracy; however, few LIBS studies employ this approach. Data fusion, although promising, is underutilized due to its high cost and low analytical frequency. This study compares calibration models—partial least squares (PLS), principal component regression (PCR), error covariance penalized regression (ECPR), and maximum likelihood principal component regression (MLPCR)—for analyzing Ca, K, and Mg in non-conventional food plants. Two LIBS instruments, featuring CCD and ICCD detectors, were evaluated individually and through data fusion. ECPR and MLPCR outperformed conventional methods, with ECPR showing superior results. Detection limits ranged from 0.003 g 100g⁻¹ (Mg) to 0.2 g 100g⁻¹ (K), and sensitivity varied between 1.12 and 12.65 (signal area)(g 100g⁻¹)⁻¹. Data fusion significantly improves analytical accuracy, and while cost and frequency factors should be evaluated, the benefits often justify its use for high-precision applications.</p>