Within-sample variation in cotton fiber len,gth is a major factor influencing the production and quality of yarns. The textile industry has been searching for approaches of improving the long fiber fraction and minimizing the short fiber fraction within a cotton sample to produce superior fiber and yarn quality. USTER® High Volume Instrument (HVI) has been widely used for a rapid assessment of cotton fiber length traits from a fiber bundle. However, its effectiveness for genetic studies has been questioned due to the indirect estimations of the cotton fiber traits that cannot be measured from a fiber bundle. To overcome the limits of the HVI fiber length traits, we utilized the Advanced Fiber Information System (AFIS) measuring fiber length traits directly from individual fibers based on weight or number. Comparative fiber length analyses showed AFIS provided higher sensitivity in detecting the fiber length variations within and among cotton samples than HVI. The weight-based AFIS length traits were strongly correlated with the corresponding HVI lengths, whereas the number-based AFIS mean length showed a relatively weaker correlation with the HVI lengths. Integrations of the weight based-length traits with genome-wide association studies (GWAS) enabled classifying the QTLs specifically associated with long, mean, or short fiber length traits and identified a false positive associated with the indirectly estimated HVI short fiber trait. Unlike the weight based-AFIS length traits, the number-based AFIS length trait did not show a negative correlation with a weight related-HVI property, and identified a single QTL that was not detected by the corresponding HVI trait. These results suggested that integrating the AFIS method with GWAS helped discoveries of the genome loci involved in the within-sample variation in cotton fiber length and characterizations of the fiber length QTLs.