General comment: all tests have confidence intervals (standard error of measurement). The difference between 135 and 130 is well within standard error for most tests. And that's for administrations of the same test. When you consider the differences between tests, the standard error would be even more generous. I wouldn't put any weight at all on the difference in total achievement. The difference between 98%ile and 99.8%ile borders on significance. It's possible there is a genuine difference, but also not. Were they tested at the same time, or some years apart? The second test administered might be expected to have slightly higher scores, due to the Flynn effect.
As to the WJ and WIAT themselves: I prefer the WIAT, based on the subtests it has, especially the reading comprehension, oral reading fluency, and written expression subtests. The WJ has the advantage of fluency measures for reading, writing, and mathematics (WIAT only has reading and mathematics), which can be important for a diagnosis of dysgraphia (due to the impact of fine motor speed). OTOH, the WJ has a much higher ceiling, as it was normed through the ninth decade of life. (Guessing it was a pretty small and skewed sample, though!) The WIAT is normed through post-secondary, but really only has material up through the 12th grade/first-year college. For a six-year-old, the ceiling should not be a significant factor (and if it is, I think you should have plenty of ammo for acceleration!), with the following exception. One issue that does come up with the WIAT with very high-functioning kids is that there may be underestimates of ability on certain subtests, because of the item sets, which are administered based on age/grade (although a savvy examiner might select a set based on an estimate of skill level). For example, if your six-year-old is finishing first grade, but functioning at a fourth grade level in reading, she will still be administered the first grade item set, which is designed to collect meaningful normative information on the majority of first-graders, including those functioning +/- 2SDs from the mean. Except that your 6 yo might be +3 SD, so she may bang her (figurative) head on the ceiling of the item set, resulting in a probable low estimate of her actual reading comprehension skills. If she has done so, the examiner can determine if that is the case by how closely she approached the max raw score on the subtest. The subtests with this kind of age/grade-determined item set are reading comprehension and oral reading fluency.
The WJ also has a number of additional subtests that can be useful for additional probing, such as spelling of sounds (spelling nonsense words, which exposes encoding weaknesses in individuals who have memorized spelling words without a strong grasp of the phonetic system). However, these subtests are rarely administered in standard batteries.
