RTFM.
The ADXL345 samples at a user-specified data rate, up to 3200 Hz. Data is stored in a fifo.
I think this system might have several problems.
First, the accelerometer output will probably have a LOT of noise on it. Generally, you average samples to eliminate the noise.
Second, assuming your 4-blade propeller runs at 2000 RPM (33.3 RPS), you'll have 132 peaks/second (1 peak/blade). With a sample rate of 3200Hz, that gives you a resolution of about 15 degrees. (132/3200) x 360=14.8 degrees. Is that good enough?