1. A bell curve structure. As you said, most players are at least average-ish in ability. Of course, what's "average" in an ability also depends on position (linemen, on average, are stronger than wideouts, while wideouts are faster, on average, than linemen or fullbacks). This means each position occupies different places. For example, WO speed might range from 75-99 with 85 as average WO SPD. Meanwhile, 70 for a DE might be average SPD with a range of 50 to 85 for the speed freaks or fast 3-4 OLB types who play with their hand on the ground in a 4-3.
2. Elite and scrubs are significantly above/below average. If average range is 60-70, then elites are 90 and the losers are 40.
3. Use the MAJORITY of the scale. Madden seems scared to use anything below 75, except for areas the players have no skill in. Of course, that brings up the "why show throwing for a DE" questions (which have merit) but that's another topic. The scale should be absolute while positional location on the scale for each ability is where relative ability comes in. So seeing a WR with 50 STR might be average for WR, but in terms of ranking this guy on a pure strength scale, he's pretty bad. However, a 80 STR for a NT might be average - but on an absolute level, he's pretty strong.
4. Stretched out ratings should indicate skill sets even in positional ratings. This goes beyond putting 12s on DTs throwing or pass blocking. It includes differentiating between a Ngata and a Freeny. Ngata is slower and stronger, maybe he has 60 SPD, 90 STR, 75 ACC, 65 AGI, while Freeny is quick and nimble, using positional advantage (due to speed and explosive first step and the ability to run his preffered arc) over raw power - 80 SPD, 70 STR, 95 ACC, 85 AGI.
This goes for OL too - a pass blocking tackle that could hang with Freeny: 75 SPD, 75 STR, 95 ACC, 85 AGI, 75 PBS, 90 PBF, but this guy might get run over by Ngata on run assignments - since his strength will matter more as would his RBS, which should be lower. Likewise this OT would get hampered by a bull rusher, since his STR/PBS aren't that hot. So it would bring up real skill set types that bring strength/weaknesses to rosters and finding matchups a) possible, and b) meaningful, if not crucial)
To me, that's what stretched out ratings could do to really help the game and what they mean to me. None of this, btw, has anything to do with OVR. I couldn't care less what happens to OVR. If there just HAS to be some "OVR" rating, take the average of the position's top 4 or 5 needed skills and be done with it, none of this weighted stuff that sometimes ignores completely various skills that actually matter on the field, or even in simmed results.