Ah! I missed that part about going back from LAB to RGB. If your working space is Adobe RGB(1998) that is the color space the image should be in after converting the LAB version back to RGB. I just confirmed that by opening an sRGB JPEG in Photoshop CS 5 (PsCS5) converting it to LAB andthen converting it back into RGB. Once back in RGB the profile for the photo was in the working space iuse for PsCS5, Pro Photo RGB.
As for why go to LAB in the first place, I have been runnign some experiments following Lee Varis' multilayer and multip le document copies 10 Channel Color Workflow where you first work on an iamge in RGB to set tonal relationships, make a flattened copy of the document and convert the duplicate to LAB, adjust color via the A and B channels (again using multiple layers), make a top layer o f the LAB version which contains the changes i nthe eaprevious LAB layers and add that layer asa new top layer in the layer stack in the original RGB docment , make a second duplicate which you convert from RGB to CMYK, and then work withthe K (black channel) to get rid of the over saturation in certain colors, make a new top layer con tainignthe changes in the lower CMYK layers and copy that top layer back as a new top layer in the original RGB document.
The workflow is a little tricky and for some (not many actually)in my canon of work it is a useful technique for controlling the shape and feel of the image. it is not something I can recommend doing with every image and practice (i.e. screwing up until you get a feel for what you are doing) is necessary. If you want to explore it check out go to: http://varis.com/video-tutorials/
I need to point out that I recently tried this 10 (really 6: Red + Green + Blue + A + B (in LAB) + K (black) in CMYK) on a Texas Hill Country landscape but ended up with better ( more visually appealing ) color by trying different camera calibration "profiles" in Lightroom 3.6 and just working with a single RGB document when post processing in PsCS5.