It should be clear by now, but the index stores the address of an entry, as opposed to the address of a response body.
You may also be interested in compiling dump_cache (src/net/tools from a chromium checkout). That tool will decode interesting parts of the headers and may help you understand how all the parts fit together.
Information for the new format is linked from the same page William mentioned... and the docs will be updated as appropriate as we go.