Each Linux server shows the identical picture, but from z/VM we know this picture can not be true. Whenever analyzing data from the Linux perspective, it WILL be incorrect. You just don't know how much it is incorrect.
Given incorrect data from Linux, capacity planning would be often grossly incorrect and incorrect to some extent even in the best case. If accounting for Linux is something that would be required, the Linux data could not be used. So what is the solution?