Double Ken's calculation for a safety margin that should work, then double the flag width if possible. The actual on time of the flag at high speed can't be guesstimated very well by rolling it past it nice and slow making pencil marks. Distance is killer with inductive proxes. Get it as close to the target as possible for as long as possible and make it as sudden a transition as possible to fully blocked. I have used 1mm feeler gauges to set the gap on a solid mount to improve this type of arrangement where the sensing range is supposed to be 5mm. Closer is better. A bigger flag with a consistent gap is always better too.
If you can put a scope on the signal you can prove that it is in or out of the signal conditioner spec. We used Wilkerson Instruments for this sort of application with excellent accuracy, if you want to try a different signal converter.
You will likely find very short on duration periods. Consider, if the off pulses were just as short, find that frequency and compare with the converter rating. Note the off and on switching speeds of the device if specified, sometimes the sensor is your high freq. bottleneck by having a slightly longer off switching speed.
If the minimum pulse width is too short or sensor speed too slow, you will have issues as you describe (missing pulses at high speed).
If the sensor is fast enough but the on time is just too quick, the problem can sometimes be overcome with a "pulse stretcher". (A Pulse Stretcher amounts to a hardwired solid state off delay timer.)
If you use one, you may get away with a single flag until you get way slow...But it can even out the on to off ratio for a specific speed and keep the on pulses above the signal converter rating, extending the upper end speed range of the system.
Sometimes, weight is an issue for the flag size, and hall effect sensor works better with a small flag.
Sometimes switching flag materials is the ticket. Higher ferrous content materials are easier to pick up with standard inductive proximity switches.
In a pinch, I like to have a nice programmable laser PNP photocell and some $100/ft stickers laying around. Have it mounted, plug it in, set the off delay timer, apply reflective sticker to about 40% of the flag frequency, in just a few minutes.
Prepare the surface well to set the adhesive well if you use peel and stick polarized retro-reflective material. Polarized is important near shiny metals or volatile ambient lighting. I clean the surface with sandpaper if necessary, crosshatch if possible, alcohol, (primer if called for).
Do this for high speed and/or semi-permanent installation. It weighs practically nothing and overcomes signal pulse width issues and distance to sensor problems.
EDIT: Keep two targets, looking at pic., photocell and million dollar tape on the orange part (throw away the flags) would be sweet. Sensor must be fast though will co$t extra for 100 microsecond speed or anything faster than 250 in my experience.
http://www.bannerengineering.com/en...oelectric-Sensors/67/PicoDot-PD-Series/#specs