I don't think that is quite correct.
Remember I said "often" not "always". Because sometimes the director intends for a sound to be localized, if briefly.
The idea of surround sound is to mimic what we hear in large spaces. Often, as much as 1/2 of what we hear is actually reflected sound, which is not localized/directional. It is sound that is reflected from a widely dispersed area and delayed from the original souce (depending on distance and what it is reflected off), that queues us as to what our surrounding are.
A good example of this would be the scene in Casino Royale where Bond steps out of the lobby of the casino onto the street. On my system, the traffic sound was comming from all around (front, back, sides, while the dialog was comming from in front of me. It sounded like I was sitting in the middle of the street, watching the scene.
In order to achieve this illusion, the surrounds need to distribute the sound they produce widely in the room and to blend in transparently with the sound comming from the mains (some of the traffic sounds would have been coming from these speakers as well).
Surrounds with drivers firing in multiple directions seem to do a better job of this.
Surround sound can mimic the 'sound' of any space, but one must remember that the reflections in a small room are not like those of a larger one; a defuse sound field (or like you said "reflected sound, which is not localized/directional"), as it exists in a large space, does not exist in a (relatively) small room. The distances are too small, the delays are too short and the reflections are higher in level.
Therefore, the sense of envelopment created by a mulitichannel soundtrack is done by the recording engineer, it is up to them to provide appropriately delayed, level reduced sounds in the appropriate channels to create the illusion, and that illusion can mostly mask the reflections from the room. The direct sound from the surround channels is, in effect, standing in for the reflections that would happen in the space that the filmmaker has put on the screen, to paraphrase Floyd Toole; the surround illusion can be created at will, without any help from the room.
Speakers need to do certain things, and be in certain places to excel when producing the desired effect. Certain places because the angle a sound arrives from is important with regard to envelopment, and do certain things because directivity is important to sound quality, the kind of 'dipoles' marketed as surrounds have directivity issues. It is noteworthy that the recommendation for dipole surrounds comes from a time when there was only a mono surround channel.
If I seem overly critical, remember that I am discussing the technical merits two methods, the enjoyment received from either is not in question.