As to your question, it seems like you could think in terms of quality standards, or something along those lines. What I mean is, suppose variable X has percentages (20, 5) respectively according to how you listed them, and Y = (90, 40). Now, if we considered 1 percentage point of improvement the same no matter what, we would have to devote time to Y simply because it has the higher weight. What I mean is, implicitly you are considering that a difference from 20% to 21% is more important than a difference from 90% to 91%, because 20% represents a failure in meeting quality standards.
As to how to change the model to reflect such ideas, I'm not sure, but anyway I think it's an interesting application.
But if I had to follow your model exactly as presented, my guess would be to devote time to the property that has highest weight always, until it reaches 100%, then move on to the one with next-highest weight, etc.