It matters a lot - any error in a read cannot be corrected for if it's the only read covering a loci. Coverage is also uneven; an average of 1x would mean a significant portion of the genome has no coverage. Even 30x would leave gaps (not sure how many as don't work with humans).
Correct - the coverage number is an average across the sequenced parts of the genome. In some areas the coverage will be much higher, in others much lower.
Importantly, what is commonly called 'whole genome sequencing' is not really that. There remains ~5-8% of the genome that is (almost) impossible to sequence with current technology, and as such has no coverage. Areas with lots of repeats, centromeres, etc.
