Lean meat percentage (LMP), a vital production trait in the swine industry, has been decreasing in the Czech pig population. To provide better control over LMP, its component traits of backfat thickness (BFT) and loin muscle depth (LMD) could be evaluated separately. Therefore, a three-trait animal model has previously been designed. With this study, we include DNA microarrays in the analysis. Data from 150,063 pigs were collected, including 1,505 individuals genotyped by Illumina 60k or GeneSeek 50k array. PLINK was applied for the quality control of SNP data. Breeding values were predicted using conventional best linear unbiased prediction (BLUP) and single-step genomic BLUP (ssGBLUP) in the BLUPF90 family of programs. Various blending, tuning, and scaling options were considered for constructing the realized relationship matrix (H). The results were subsequently compared under the linear regression method. Overall, ssGBLUP results showed higher bias, higher dispersion, and lower population accuracy. However, individual reliabilities of young cohorts increased by up to 10%. This discrepancy might indicate unsatisfactory quality or quantity of SNP data or an unsuitable setup of the H matrix.