Sorry I missed your last message. I'd be interested in helping to test if it's not too involved, for example if you have a beta version of your samples application
I tested this in the latest ArcGISRuntime samples in version 100.13 and the problem is the same. I retried all the files I showed previously, and every problem I discussed is still here.
To quickly summarize (these are new examples from 100.13):
1. In Dips mode, nothing is anchored correctly. If you zoom in far enough, the model will always diverge from the actual point in some way. Here you see a Dips cube and a Meters sphere (anchor point is Center). The cube is sticking out the bottom of the sphere when it should be centered in the same place as the sphere.
Here, the anchor point is Bottom, and you can clearly see that the model is not attached to the anchor point, which is given the same point as the Meters-based sphere:
Here it's anchored to Top, and you can see daylight between the actual anchor point and the model:
2. Origin and Bottom appear to be nearly the same in Dips and Meters modes. I have seen them diverge, but rarely, and Origin has nothing to do with the origin of the model under any circumstances in any model format I've tried (and I tried a lot of formats). Here's a cube which is centered at 0,0,0 in Dips:
And here it is in Meters:
Both are incorrect in the same way. No point of origin is respected in any of the tests I did, moving the origin around a cube model. The Origin anchor point definitely does not anchor at the origin, it's still similar to or the same as Bottom, and the Center is still using an algorithm that's difficult to figure out. It seems to be maybe a weighted average of face area, I think. Vertices that aren't part of a face are not considered, I know that much.
3. Much less important, but I can't do your SimpleMarkerSceneSymbols in Dips mode. It would be very nice to be able to create a cone, sphere or cube in Dips mode. I assume this is possible but was missed in the API or something.
Generally speaking the SimpleMarkerSceneSymbols make sense in terms of how they work (though I'm still confused about how Center is calculated). Here I plotted them in a similar way to the other models:
Their anchor points seem to work as desired, generally speaking. Their Origin anchor points clearly work differently than all other loaded models, since they seem to actually be in the center of the models, and I haven't seen any other case where I could get that to happen. If I could get the same behavior for loaded models I'd be very happy, though I can't test their behavior in Dips mode.
4. Minor issue, I wish I could load models from a pack Uri. If it doesn't work for some reason, that's fine, it's just strange using Uris but not being able to do it the same as I would in XAML to load a local resource file, I have to write the file to disk and load it that way.
The runtime performance of 3D appears to be better in 100.13, but from a quick test it still seemed slower than in 100.6. I haven't benchmarked it though, and it definitely appears to be improved from other versions between 100.6 and 100.13.
I just realized I could orient my meshes vertically, since Top and Bottom are more consistent in their behavior, so I'm going to go try that, but it isn't a real solution for models that aren't symmetrical (since I don't know how the other axes are averaged), and even in Top and Bottom mode for Dips, the models don't really stay attached to the anchor point.
Let me know if you may be able to fix this or if you can at least tell me how the centers are calculated. I can give you the sample code and models I used to do this testing if it would help you, it's similar to what I already provided in this thread.