Whether diffused or non-diffused sources are better or not depends on the texture of the object the light hits and with the goals of the shot. Specular reflections on skin in conventional portraits isn't desirable which is way large sources used close are the best tool for the job. But if shooting a body builder and going for a chiseled look the opposite wound be the case. You'd want more parallel rays from the lights creating specular highlights, which has the net effect of increasing the contrast range from shadow> highlight by making the highlights brighter. The brain translates that steeper contrast gradient into the impression of steeper, more angular 3D shape when seen in the 2D image.
Collimated (parallel light ray) sources are better for animals (stuffed or real) due to the more specular refections they create on the hair shafts. It's the sparkle of the catchlights that give fur it's 3D texture. It is more pronounced on real hair because real hair seen microscopically consists of flat scale-like plates which catch and reflect light back to the lens. What happens with larger more diffuse sources is that the reflections off the hairs scatter in directions other than the lens axis so they highlights appear less specular.
Stop and consider how 3D shape is defined in a 2D image with contrast gradients:
One way to affect the overall shadow > highlight gradient is with fill and shadow tone, making the shadows lighter or softer looking. The other is to making the highlights more or less specular, clues the brain uses to determine whether the surface texture is smooth and or matte finish.
Much is made about the ability of larger sources used close to 'wrap' the light around 3D objects but in most studio settings the 'wrap around' effect is caused mostly by the fact the large light footprint overlaps on to the walls and ceiling creating a skylight-like fill effect. Even when a small single key light is use light bouncing off the opposite wall will create a significant amount of "spill fill".
For example, of you still have your bear set up go to the back of the room and shoot a wide shot like this one... http://super.nova.org/TP/CEG_Oblique_Setup.jpg
you will be able to visualize how much spill fill is bouncing off the ceiling and walls contributing to the "wrap" in the shadow transitions you are seeing in your test shots. One of the variables when adding a diffusion panel is a wider overall footprint and more 'spill fill" off the ceiling adding to the wrap-around fill effect. Do the same test at night in your driveway were there is no spill fill and you will see different results.
Again that's not to say one tool is better than the other, more that how you conduct your tests affects the outcome and you need to take variables like the texture of the objects and spill fill into effect to make heads or tails out of the results seen. Erroneous conclusions can be reached if all the variables aren't taken into consideration. If your goal is to see the different in diffusion vs. non-diffused PLM at night in the driveway where there is no spill fill variable or ambient light would a more objective test protocol.
At night outdoors on a non reflective surface like the driveway with a single light in a short lit / oblique view (light 45° to the side of the nose at a downward 45° angle to the eye line) you will objectively see the true "wrapping" characteristics of the source without the wrapping effect of the indoor spill fill or the skylight during a daylight test. If you repeat the same test protocol (subject, light distance and angle to face) doing the day and indoors and compare the results of all three you'll be able see what is "wrap" effect due to source size and what is due to the ambient or bounced spill. I've done tests like that many times over the years to gain a better understanding of the cause and effect of modifer size and how the results change with modification and distance.
.