I work for a company that makes airflow measurement devices and to do it accurately in high temps is expensive. We use inconel or hastelloy probes (for high temps - up to about 1900 deg F) that utilize Pitot Averaging technology in conjunction with ultra-low dp transmitters. For a system that is set up correctly you can expect 2-3% accuracy. The problem is that you need a measurement location that has a uniform air velocity profile. This is easier with small ducts, but you still need about 4 diameters of straight duct run upstream of the measurement location. Also you need to consider the losses with venturi type devices.
The more I study these rocket stoves and other similar rocket stove type applications, the more I am convinced that the simpler approaches give the most bang for the buck.