I did something once that turned out pretty interesting.... Now, it won't explain why pressure/wind relationships aren't always exactly constant, but it can give you a feel for the amount of variation. Windspeed depends on pressure right... Take a storm (any storm, but preferably one that had large variations in windspeed, so like a stronger hurricane).... Plot pressure versus windspeed in Excel, where pressure is the "x" and windspeed is the "y". Add a trendline to the plot (a regression line, really.. Excel just calls it a trendline) and display the equation and "r-squared" value on the graph. It'll generally be a pretty good relationship and while the slopes (which show the relationship between pressure and windspeed) will always be in the ballpark... there is a fair amount of variation.

Something else to do (I tell you, I'm the biggest nerd ever, and I freely admit this) is take a large sample of pressure/windspeed data from many, many storms and do the same thing.... You should end up with the relationship that is the overall estimate between pressure and windspeed that various sites use... I've never done this though.

For what it's worth...

Edit: And the windspeed it what counts as far as Saffir-Simpson numbers, but (as discussed on that thread... I think that's where) damage comes from more than just wind... in fact, storm surge is responsible for most hurricane-related deaths, and flooding is a whole other damage issue.... I'm still all for a Modified-Mercalli-type scale for hurricanes that takes everything into account and provides a gauge of damage.