I'm not sure why the test publishers are so insistent on keeping subtests that are poor measures of 'g'.
Because, according to a study I read and if I understand your question, IQ tests are generally used in an academic setting to determine school placement. This study showed that FSIQ, not GAI, correlated better with longterm school success. Those things like processing speed and working memory, which have little to do with how intelligent a person is, have very much to do with how successful a person is in school (notice it correlates with how well a person does in school, not how successful they are in their life acheivements). I think low PSI and WMI scores are often discounted with no thought given as to how they might negatively affect a child in an advanced setting. A huge discrepancy may not have any effect on a child's ability to be the world's next great thinker, but might very well effect how they get through school. More than a few of histories great minds were washouts in the classroom.