A weak palindromic nucleotide motif is the hallmark of retroviral integration site alignments. Previously, the motifs were explained by an overlap of the non-palindromic motif being present on one of the half-site of targeted sequences. Here, we applied multicomponent mixture models to integration site sequences of diverse retroviruses. We demonstrate that the weak palindromic motifs result from a combination of independent sub-motifs restricted to only a few positions proximal to the site of integration. The sub-motifs are formed by either palindrome-forming nucleotide preference or nucleotide exclusion. Using the mixture models, we also identified HIV-1-favored palindromic sequences in Alu repeats serving as hotspots for integration. Our work presents a novel statistical approach to the analysis of retroviral integration site sequences, which can form a valuable tool in the analysis of DNA motifs. The presented results shed new light on the selection of target site sequences for retroviral integration.