Firstly, you are still only manipulating what the cameras sensor captured in one exposure. You are not adding any kind of depth which wasn't there before, like you would with multiple exposures.
You are not adding anything to the depth of the image from the camera, obviously, but what you are doing by using multiple raw developments is allowing yourself to transfer more of the range that the camera
did record to the program dealing with the image. Each time you 'develop' a raw image you are 'clipping' a 12 or 14 bit sensor image (with modern DSLR's) to the color space gamut required for JPeg or TIFF formats. As you can effectively select different parts of the sensor gamut, if you develop with +2 and -2 stops compensation you will end up with three photgraphs having, between them, the same shadow and highlight detail as if you had taken 3 separate exposures at, say, 1/125, 1/500 and 1/2000 (constant ISO and aperture). *
If these are appropriately combined you will end up with an HDR image. A real HDR image that has a 50% greater range than a single developed RAW image.
Sure, you can take an image of a house on a bright day, tweak the RAW exposure to create an 'HDR' but ultimately you can only do this because the scene itself doesn't have a big range, and it can make it look pretty.
You can do more than that!
You can take a picture of a house where the higlights would wash out and the shadows would lack detail in a singley developed RAW image and produce a real HDR which can then be tone mapped to show detail in the shadows and highlights
that could never be extracted from a singly developed image.
It's very useful for creating good, natural, 'non-HDR' HDR images.
This is where you are confusing yourself....
Well, I didn't confuse myself but I must admit that it was a stupid way to express what I intended which as that it was a good technique for producing a 'natural' looking image as opposed to one of those very obviously generated by tone mapping a very high dynamic range image (i.e. more than +/- 2 stops from the central exposure).
* It is very easy to prove this for yourself.
Select an 'impossible' subject where you know that you will get burned out highlights and flat black shadows. Expose to balance the lack of shadow and highlight detail.
Now develop the RAW image three times using 0, -2 and +2 stops compensation.
Examine the results.
You will note that in one you can see shadow detail, in one you can see highlight detail and in one you can see neither.
These can be programatically combined into a non displayable true HDR image that has a greater bit depth for each colour and that image can then be tone mapped to print or display on a screen.