A note on using 2-3 decimal points as an accuracy metric or for that matter any figure of merit. Assume I make a measurement of a 1.0 centimeter this is 0.00010 kilometers. The first result has one decimal place and the second has 5, is the measurement any different by only changing the coordinate system?
You need to be a little more quantitative when making requirements on accuracy.
In my research on weather sensors it seems the humidity is hard to measure to within say +/-3% without spending a lot of money and doing occasional instrument calibration. Buying the equipment for calibration will drive the cost up very quickly.
When starting out on a design of a sensor system making it very accurate is always the initial response. Somewhere along the design process you need to ask yourself what it is you need to accomplish with the instrument and let that drive your accuracy requirements. In my first reply I talked about using a humidity sensor that was accurate to +/-5% and saying that was good enough for my application, density altitude measurement. Humidity's influence on the density is very small so I could get away with a lower quality measurement of it, ther +/-5% error on the humidity did not map over to a +/-5% on the density altitude.
On to sensors, for measuring pressure Freescale was a great line of MEMs sensor and I would encourage you to look at their website. MEMs sensors was dramatically driven the cost and size of the sensors down and you probably want to exploit that.
wade