Hmm, I'd say ask your assembler... likely they'll play with the mask to suit their process anyway, but you can always see what's typical.
I don't think it matters very much; most devices are designed for that sort of thing, for full pad coverage that is. I think the most sensitive would be a full pad under a QFN, which could float up, whereas the leads usually have some toe fillet*, meaning they'll want to sit lower rather than riding up on a blob. (BGAs of course don't matter much because they have a huge ball, you mainly just use the paste to be sure it touches the pads.) And as for placement, it's not a big deal because the paste flows pretty well on heating and melting.
*Found a doozy recently. FT232RQ does NOT have traditional QFN construction. It's built slightly more like an LGA (no toe). Also, their drawings of the part stink!
I know there are differences between types of masks (material, thickness, cutting method, taper, surface finish) and geometries (paste tends to be retained in sharp corners, versus releasing better from round corners), but as for what works best... there's probably a broad range of "good to great" ways of doing it, so that asking for one best way is kind of pointless.
But yeah, I just draw the things, I don't manufacture them. If there's an assembler around, maybe we can get some more specific answers?...
Tim