AWR1642BOOST: The magnitude ratio between the original signal and the FFT data

Firstly, note the documentation which is using latex to graphically show the formula, when you write it as text is N/ceiling( 2^[log4(N)-1] ), you did not write ceiling so just pointing that out. As you pointed out, the FFT definition would have scaled only by N, the additional reduction by a factor of 1/ceiling(..) is because the DSPLIB FFT function performs an internal scale down between stages to prevent intermediate overflows in the computations as documented in the user guide at  http://www.ti.com/lit/ug/spru565b/spru565b.pdf 

"Since DSP_fft16x16r and DSP_fft32x32s performs scaling by 2 at each radix-4 stage, the input data must be scaled by 2^ (log2(nx)−cei[log4(nx)−1]) to completely prevent overflow"

The document is not crisp about "r" (rounding) versus the non-r (which we use) versions, but both have the same scaling. The DSPLIB is missing this scaling documentation in the API documentation (.chm) that is shipped in the release itself which will be updated in a future release. If you want to check this yourself, you can create a stand-alone test that calls the dsplib functions with a tone with known amplitude (sufficiently low so as not to overflow) and check the output amplitude to find out the scaling factor. If you find different that what I mentioned, you can report back here.