Can you take a couple of screenshots of your workflow please? I want to make sure I understand your process to try to isolate the Photoshop part from the LR part.
Some questions. If you start your merge with your three source images (or however many) from within the merge to HDR Pro dialog in Photoshop (i.e., do not use LR at all in the process) do you get the same results? Do not use Bridge or LR, just use Photoshop - File > Merge to HDR Pro and select the source files in the dialog that pops up. In that merge, do not adjust the White Point Preview slider - just select the 32 bit option and complete the merge. The newly created file should appear in Photoshop in 32 bit per channel mode, in the linear version of the raw conversion utility's color space.
When you use PS to make a Merge to HDR Pro, the file browse dialog will prompt you to pick the raw images you want to merge - navigate to the working directory with your source raws and choose them. If these source files have XMP sidecar files next to them that contain raw edit information, move those for now, or make copies of the source raws and put them in a new directory and use that as your working directory. In other words, let's get rid of all of the possible confounding issues related to raw conversion that may be causing problems.
The images will be imported into PS and converted to 16bit rendered files - if you set up the information bar at the bottom of the display to show the document's color profile information, you will see that each imported and converted source image is a 16bit image with a particular color profile (e.g., ProPhoto RBG, etc.). Eventually, when all of the source images have been imported, the Merge to HDR Pro preview window will appear - make sure you select the 32 bit option and do not adjust the White Point Preview slider, just accept the merge and create the 32 bit file in PS.
The newly created 32 bit file should have the linear version of the color space of the source images. The newly created file should appear just like it did in the Merge to HDR Pro preview. Now is the time to check the View > 32 bit Preview Options menu item to see what exposure PS has chosen for your preview. It may not be "0", regardless of whether or not you have adjusted the White Point Preview in the Merge to HDR Pro Preview. Now, if you set the 32 bit Preview Options > Exposure to "0", I am guessing that the newly created HDR image will appear dark. As far as the heavily saturated and over contrasty, I suppose this is a function of what subject you are shooting, or if there are adjustments you made to the raw files in LR that are affecting the HDR merge.
Here is an example workflow in case my description was vague:
1) Use Merge to HDR Pro in the Automate menu of Photoshop to select your raw source files and bring them into PS. Here is the preview window during the merge:
Use the 32bit option, do not manipulate the White Point Preview. Note the appearance of the preview - if it is a true HDR image, parts will be blown out or blocked up in the preview.
2) Click OK to start the merge and open the newly formed HDR image in PS. Here is a screenshot of the merge result:
Note that the new HDR image has a linearized color profile of the source image profile. This is the 32 bit HDR image, ready for toning and conversion to 16 bits.
3) Take note of the appearance of the newly created HDR image, in terms of the way the image is rendered in PS. It should appear identical to the preview in the Merge To HDR Pro preview as shown in (1) above. Now, check the View > 32 bit Preview Options dialog to see how Photoshop chose to display the preview for you:
Note that Photoshop chose to boost the exposure 5EV or so for the preview in PS. This can be significantly misleading in terms of what you are expecting from the HDR data.
4) Change the exposure here to "0" - note the change in appearance of the image:
Now you are cooking with gas. However, this rendering is still a preview, adapting the 32bit data to your LDR display. You may see some artifacting, banding or weirdness.
5) If you would like, you can tonemap the 32bit image in PS to get the HDR data compressed and redistributed in preparation for conversion to LDR (16bit). Go to Image > Adjustments > HDR Toning:
to bring up the PS toning panel.
Here you can apply PS's Local Adaptation tonemapper, or experiment with the other tonal range compression algorithms. Make your adjustments to get all of that HDR data to look the way you want it to look in LDR land and hit "OK" to accept the adjustments.
NOTE: the file is still 32 bits per channel, its mode has not changed to 16bit yet. This permits you to continue to make 32bit adjustments if you choose, without dropping all of the HDR data until you are ready to convert to 16bit.
6) When you are satisfied, you can change the mode of the image to 16bit. This will bring the HDR Toning dialog again, but, since you have already done your tonal range compression in step 5, all you need to do is a straight mode conversion with Exposure and Gamma - Exposure set to "0" and Gamma set to "1" (unless you want to change these settings).
Now you should have a 16bit file that has the HDR tonal range compressed and mapped the way you want it. You will find that, during the toning and conversion to 16bit, the color profile of your image has probably changed to the working space you use in PS. No biggie. This 16bit file is ready to be imported into LR and the file should appear as it does in PS.
LR, as of 4.1 or 4.2 can accept 32bit per channel TIFFs - that is, if you choose to use LR to perform tonemapping on your HDR data, you can, if you save the 32bit file in PS as a 32bit TIFF.
I do not use LR. However, if you insist on prepping your raw files in LR before merging to HDR, be aware of what you do in the raw conversion process. Christian Bloch has a nice section in his new "HDRI Handbook 2.0" about preprocessing raw files in LR for merge to HDR in Photoshop or other applications. If you are using LR4, to create a flat raw conversion for HDR merging, follow this advice (from p. 184 of his new book):
1. Select the best (middle) exposure for your reference.
2. Set the Exposure to -1EV and Contrast to -40.
3. Set the white balance.
4. Use low sharpening as as much noise reduction as necessary.
5. Enable lens correction by profile only if you really need to correct geometric distortion or vignetting.
6. Remove chromatic aberration should always be enabled. It's now just a single checkbox - it's no longer based on camera profiles, and there is not hand crank anymore.
7. Sync all settings to all other images.
8. Export 16bit TIFFs or launch your favorite HDR plug-in.
NB: You should zero everything and set the tone curve to linear before performing the above steps.
The adjustment to Exposure/Contrast is based on Christian's experience with flat conversions in LR3 and trying to get the new PV2012 in LR4 to match the flat conversion of LR3 (PV2010).
See if eliminating LR from the equation presents you with the same problems you are experiencing. LR is unnecessary unless there are lens corrections or other raw adjustments you think you might need as a pre-process. If that is the case, you need to create a flat raw conversion or else colors and tonal information will not be accurately rendered for HDR merger.
Start with the raw files outside of LR and do a merge in PS and see if that eliminates some of the problems. Post back here! We'll get this working.