For this study, a double haploid population of Japanese flounder was produced using mitotic gynogenetics. Body weight and 7 morphological traits (i.e., total length, dorsal-fin length, pelvic-fin length, body height, caudal peduncle width, head width and trunk length) were measured. After normalizing the measures, we acquired a data set that could explain 90.4% of these principal-component analysis traits (PPCTs), and accordingly we denoted these PPCTs. Next, by modifying conditional probabilities of quantitative trait loci (QTL) genotypes on the distorted flanking markers, Bayesian model sel­ection was used to dissect the genetic arc­hitecture of the PPCTs with four genetic linkage maps, created with the software JoinMap4, MapDisto, JoinMap-DistortedMap and MapDisto-DistortedMap. The different methods used to construct the genetic maps and QTL mapping of the PPCTs produced variable results. Comparing the map produced with JoinMap4, the rel­ative positions and the genetic map distances of the partial separation markers differed from the map produced with MapDisto. Furthermore, five separation markers were not located at the corresponding linkage groups. Comparing the maps based on JoinMap 4 and MapDisto and after correction by DistortedMap, the positions were not changed, but the genetic map distances of the partial separation markers were changed. In the overall maps, three additive-effect QTL were detected in linkage groups 6, 9 and 22, of which a negative effect could account for 12.95%, 14.85%, 11.56% and 11.76% in linkage group 6; a negative effect could account for 13.86%, 13.27%, 11.17% and 11.25% in linkage group 9; and a positive effect could account for 5.68%, 4.36%, 4.97% and 3.58% in linkage group 22. At the same time, 28, 19, 29 and 20 pairs of additive-additive interactions were identified, mainly distri­buted in linkage groups 6, 7, 9, 17, 20 and 22, and these interactions could account for approximately 2.19%-17.62%, 2.40%-22.26%, 2.08%-26.0%, and 3.16%-22.05% of the variance in the PPCTs, respectively. We believe that the results of QTL mapping were more accurate in the linkage map using DistortedMap based on JoinMap4.