"You seem to be saying some mental events don’t have physical causes, or maybe that some mental properties are not encoded physically."

I think he is only saying that an empirical language alone will not be enough to relate mental processes to physical ones. You have to define what a mental event really is before you can properly test anything about it, and depending on what definition you choose it might be inherently un-testable. If you define a mental process as being composed of physical processes from the start then of course you will conclude that all mental processes are influenced by physical ones. There are other possible definitions though such as physical processes being composed of multiple mental processes (the opposite of the last) in which case no physical test could be performed to verify a "change" occurred in the mental subject (but a "mental test" of sorts might still be able to tell us)