On the M-theory approach to (compactified) 5D field theories