Its difficult to occlude the reflection and even there might be self occlusion if the objects itself got reflective surface so we use technique called Reflection occlusion this technique was first developed during Speed II and enlisted into full time service on Star Wars: Episode 1 by ILM.
Reflection occlusion is calculated by shoting number of rays in the reflection direction, if the ray hits the object it will be black pixel and if it doesn't then it will be white. we cloud simulate this in 3d app, for example we cloud add base color(white) to objects and if the object reflects then the color won't be white. so finally we will get a reflection occlusion pass. And I guess in real shot we can just take the enviorment map and make that into
This occlusion pass is multiplied with the reflection pass in a compositing software. So we will get accurate depth in the reflections.
This technique is basically similar to ambient occlusion approach.
so our final out put will be something like this
Ambient Occlusion + Environment Maps * Reflection Occlusion (Enviorment map Multiplied with Reflection Occlusion)
Even we cloud use as HDR image as our enviorment map for more flexible and better results. Using reflection occlusion we cloud get some pretty realistic output!