When it comes to HDR, it is important to understand the true nature of what your are doing. High Dynamic Range images contain a floating-point dynamic range that is nearly infinite. No device that exists is actually truly able to render the full dynamic range offered by a 32bit HDR image. To be able to view your HDR images, you ultimately have to downconvert them to a smaller integer bit depth, such as 16bit or 8bit. When you do so, you are mapping tones from the full range provided by the 32bit floating point image to the limited range of 16bit or 8bit integer images.

Even though a 32bit HDR image can technically contain and represent an immensely wide dynamic range, its beyond visibility on any modern computer screen, and well beyond what any printer is capable of. In cases of extreme contrast, such as photographs with the sun in them that also include deep shadows, or in your case, night photographs that include artificial lighting via lamps and the like, there is simply too much contrast to compress into a 16bit or 8bit integer image or print. If you try to tone map such an image, you'll undoubtedly get things like posterization, harsh edges around highlights, undesirable color shifts, excessive noise in shadows, etc.

There may be some things you can do to improve your images when taking the shot. Providing extra lighting for the shadows is the first thing to try. Flash can be useful for filling in some light into deep shadows. You may need to bring along some extra lighting. Try following the ETTR technique, or Expose to the Right, where you overexpose your shots as much as possible without clipping highlights (or in your case, you may want to clip them a little bit because of the extreme contrast difference). This essentially dedicates more of the sensors dynamic range to capturing shadow detail. You can recover highlights in post-processing, and effectively achieve a more balanced shot. Try combining ETTR with some fill-flash as well to get the best shots possible.

You might also try Exposure Fusion, which is an alternative process to HDR. Expose a few shots, one to capture the bright highlights of street lamps and the like, and a few overexposed shots to capture as much shadow detail as you can. Using a tool like Photomatix, you can "fuse" the exposures together to merge shadow detail into shots that contain highlight detail. The result is similar to HDR, but simpler and often produces more natural results.

Finally, some of the exposure stacking tools for night sky photography might be helpful with images that have had their shadows lightened but display too much noise. Using burst mode of your camera, snap a rapid sequence of shots of the same scene (best done with a tripod/cable release to avoid changes between frames) that covers as much contrast as you possibly can in a single shot (ETTR might help here as well). Recover shadows and fill light with a tool like Lightroom, and merge a sequence of a single scene together in a photo stacking tool designed for stacking astrophotography shots. These tools excel at merging images and eliminating noise in the darker parts.

Note that for almost all of these things, it is important to work in RAW through the whole process if you can. Don't convert to TIFF or DNG, use copies of the original RAW files from your camera. This is particularly important if you take the astrophotography stacking route, as those tools have a variety of pretty advanced algorithms that work directly with bayer array data to produce the most ideal, low-noise results when stacking.