You're very welcome!
Yes, as stated above, I have no better answer other than visual inspection for the "accuracy" of the skeletonization maybe someone else does?
One explanation may be that the structures you are getting are not suited to be represented by a skeletonization.
Another explanation may be that your binary images are noisy and that this noise affects the skeletonization. There are, as far as I know, several ways to reduce noise in a binary image: you could try
BoneJ>Purify, successive dilations and erosions (
Process>Binary>Dilate/Erode) (as described e.g. in this paper), Gaussian or Median filtering (
Process>Filters>Gaussian blur 3D/Median 3D), despeckling (
Process>Noise>Despeckle) or combinations of these. I don't know which one, if any, will work well for you.
A priori I would say different experimental conditions do not mean different outcomes, but I don't know the specifics of your experiment. So I think you should keep this possibility in mind. In fact, the paper we mentioned earlier used Analyse Skeleton to show that the angles between the bone trabeculae of different people are very similar.
Yes, it is still in an experimental state - if you have the time and patience and reducing the noise as explained above doesn't help, I would recommend giving it a try, but installing it alongside your current ImageJ/Fiji version. Maybe the inter-trabecular angle algorithm helps you, maybe not. As stated in the previous post, if your distributions are similar because of noisy short edges, I think there is a chance it will help, but I don't know, unfortunately.
Hope this helps!