Microarray data should always be considered relative, regardless of normalization. For example, if a gene’s log2(intensity) is 6 in cerebellum and 5 in cortex, you can believe that it is twice as highly expressed in cerebellum, but you should not use the value of 6 in comparison with any other gene. Another example is that multiple probes for the same gene often have different log2(intensities). That said, we have published a paper on how to make the microarray data from the HBA more accurate on an absolute scale if this is important.