Power to Weight

The "10hp per tenth" only works for some cars. The power required to move the same mass faster is a function, not a constant. So at some point, and the point is determined by the weight (again assuming no wheel spin and optimised car), it will take more than 10hp to go .1 faster.
In terms of affects of weight loss, looking at bjk's post, you can see that by reducing mass at the wheels you get the most bonus. That's because if you lose say 10 lbs of sprung weight, it's 10 lbs. If you reduce 10lbs from the wheels, you effectively reduce three factors: total weight (made lighter), unsprung weight (parts can react faster), and rotational mass (less resistance to acceleration). All three are affected by the particular 10 lbs shaving ... The saying I've always heard was unsrung weight "counts twice" what sprung weight does because of the net effects.