Your statement seems true under its hypotheses, but becomes irrelevant if he assumptions are not true.
It isn't true under any circumstances, regardless of the validity of the qualifying assumptions. Bit depth limits maximum dynamic range, but does not define it. Even in purely theoretical terms, dynamic range cannot be defined by bit depth alone, the issue of acceptable shadow posterization (how many levels are required for acceptably smooth shadow tone gradations?) must be factored in to the calculation, and is going to reduce DR 3-5 stops from bit depth depending on how picky you are about smooth shadow tonality.
And of course, you cannot eliminate real-world considerations such as sensor, pre-amplifier, and ADC noise, the color channel multipliers required for proper white balance (anything other than 1:1:1 is going to decrease DR by making one of the color channels clip sooner than it would otherwise), and the issue of the ADC not outputting 100% of its numerical range.
To properly measure DR, one needs a standardized method of measuring all noise factors, weighting the negative visual impact of each, and defining an "unacceptable" threshold value for the combined result. It's easy to define the clip point, but we're still measuring the location of the noise floor with a rubber ruler. To do this appropriately, I think that the following factors need to be considered:
Chroma error: The distance between a recorded *a*b coordinate and its actual location in L*a*b space.
Luminance error: The difference between the recorded L value and its actual value in L*a*b space.
Frequency: High-frequency noise is less visually intrusive than low-frequency noise; fine grain is not as annoying as big blobs.
The fun part, of course, is defining the weighting for all these factors and devising a formula to calculate a noise measurement that accurately defines how intrusive any given set of noise factors is on a given image, and then defining a threshold noise value that could be agreed on as the noise floor. Then there's the issue of defining the correct methodology for capturing the image(s) that can be analyzed to accurately calculate all of this stuff...