I have found the following interesting reference "The ISO Definition of the
Dynamic Range of a Digital Still Camera" which, at pages 6, clarifies your formula.
It is based on the assumption that a useful signal is one for which the Signal-to-Noise ratio is >= 1.
This luminance level is your denominator N_D. The numerator is "the maximum luminance that receives a unique coded representation (the “saturation” luminance)", L_clip in your notation. But digital luminance is not equal to luminance (The famous "gamma correction"). This slope factor is the G_I (the incremental gain).
L_sat and L_clip are directly related: if one is known, the other follows by simply multiplying (or dividing) by a factor. It is a matter of definition.
The 1.4 factor (roughly the inverse of 70%, as you note) provides a sort of a buffer (the paper calls it the well-known so-called “half stop margin” against overexposure).
The ratio of:
- (numerator) Gamma corrected maximum luminance
- (denominator) lowest luminance level where the SNR is >= 1
gives the output, the dynamic range of the sensor. It is (as it must be) based on a series of convention: but if you apply them consistently to various sensor you can numerically describe them according to this metric.