We revisit the results of single-molecule DNA stretching experiments using a rodlike chain (RLC) model that explicitly includes some intrinsic structural disorder induced by the sequence. The investigation of artificial and real genomic sequences shows that the wormlike chain model reproduces quite well the data but with an effective bend stiffness A(eff), which underestimates the true elastic bend stiffness A, independently of the elastic twist stiffness C. Mainly dominated by the amplitude of the structural disorder, this correction seems rather insensitive to the presence of long-range correlations. This RLC model is shown to remarkably fit the experimental data for lambda-DNA when considering A approximately 70+/-10 nm (>A(eff) approximately 50 nm), in good agreement with previous experimental estimates of the "dynamic" persistent length. From the analysis of large human contigs, we speculate about the possible dependence of A(eff) and/or A upon the (G+C) content of the considered sequence.