reflections done with environment mapping usually look jagged. i don't know much about environment mapping, but from what i kno (which could be completely wrong) cubic environment mapping takes a cube-map image and according to the camera's position in relation to the model and the normals of each of the model's vertices, finds each vertex's position on the cubemap (in such a way that would simulate a reflection), and interpolates each pixel's position between vertices on the cubemap according to each pixels position in relation to the vertices making up its face. i'm terrible at explaining stuff but i'm sure you know what i mean if you think about it (and maybe after you stretch what i said slightly to sound more like how you know it actually works). the thing is these "reflections" always look stretched and jagged.
using a normal map and the normals of each face couldn't you figure out new normals for each pixel (in the same way however it works when normal mapping is used to alter a models lighting), and just colour that pixel according to a place on the cubemap (which is determined by the pixel's new normal and the camera's position in relation to the model)? i hope you know what i mean. or maybe i'm completely lost. lol soz if i'm just wasting all ur time.

julius