"he just looks at the cortex and not the oher "low level" stuff which is imho also highly important"
Highly important for what? Certainly not for understanding the fundamental principles on which the Cortical Learning Algorithm operates. Highly important for understanding the basis of much of human behavior, yes, but not for constructing a theoretical framework for machine intelligence. The theoretical equivalents of the "low level" stuff perform important implementation specific
functions. Typically these involve mediating efficient transfer and performing rudimentary preprocessing on the sensory inputs to an HTM layer, as well as potentially modulating CLA parameters. Again, these functions are not, however, inherently related to the core theoretical principles.
"you can't just take one part out of the brain (cortex) and try to understand it without looking at the interactions between different (non cortexial) brain regions [because of the emergence, if you take one critical part out it doesn't work, etc.]"
That is simply false.