4. Payloads on flying things have severe weight and volume limitations. This limits what you can deliver to things that are small and light.
5. An autonomous drone represents an opportunity for theft or interference. See the drone coming, then: steal the goods when it delivers, interfere with its GPS and force it into an error mode that lands the drone, etc. You don't even need to be at the destination.
6. Multicopters, the devices most people think of when they think drones, have terrible range and payload capacity. The flight mode is akin to helicopters, "beat the air into submission". On the upside, they can hover on the spot. In contrast, plane-style drones which fly aerodynamically or "on the wing", have more range (being more efficient), but have difficulty landing on specific spots and then returning back to base. There are solutions around this, but not 100% viable in all cases. VTOL planes exist for instance, but they're finicky. Parachutes as a delivery mechanism also exist, but not everything can be delivered by dropping a payload.
And much more...
Air delivery exists in places where these aren't concerns and where the cost benefit analysis skews the other way. Zipline [0] has been doing a great job delivering blood to remote areas in Rwanda for a while now. Their product, payload, delivery method, geographical location, regulatory environment, all align to make it worth it. Watch some youtube videos of their operation, it's pretty neat.