The first article sounds like they're trying to sell something- "The Nelson-Denny reading test was a better predictor of achievement (GPA and NBME) than the MCAT reading subset." I don't know about the first one.
The only evidence for the claim that I have come across so far is an indirect one- "Claudio Violato, PhD, and colleagues published an article reviewing performance on the MCAT exam and the Medical Council of Canada (MCC) qualifying examination parts one and two (3). Part two of the MCC qualifying examination is structured similarly to both parts of the USMLE Step 2, emphasizing clinical reasoning skills. The authors found that of the four subsections of the MCAT (verbal reasoning, physical sciences, biological sciences, and writing sample), only verbal reasoning significantly correlated with performance on the MCC qualifying examination part two; undergraduate grade point average (GPA) significantly correlated with both parts (3)."
Maybe this is why Canadian med schools have insane cutoff scores for the VR section (what is it, like 9-10?)