We seem to have a unique issue with Spectrum properly modeling a fully trunked network (no edge routers, just L3 switches on same vlan), This portion of our network consists of multiple trunked sites across a fiber backbone. As such, each switch that sits on the fiber edge is a neighbor to all the other switches that sit on the fiber ring edge. Therefore, Spectrum is not able to properly model which device is upstream / downstream for correct fault isolation. As shown in this topology model, the deice HPTvilleRehab1A is not polling as it is upstream of the other devices that are showing suppressed and modeled in Spectrum as being downstream. Yet I am not getting a down critical event as I should from the HPTvilleRehab1A switch device. As such, it is not being seen at the top of our critical event lists as we have our console sorted by severity, and no critical event for what Spectrum sees as the top device in the upstream / downstream models.
So my questions is, should this portion of our network be discovered differently as these are all trunked sites with all the devices being neighbors on the same vlan? I have already went in once and re-discovered them, and in not using Spanning-Tree Tables this time, Spectrum did not model them as all each others neighbors. But Spectrum still has a hard time determining upstream / downstream neighbors for proper fault isolation.
This is how we have the device discovery protocols set for this particular landscape: 
Any help or guidance would be appreciated as Spectrum is not accurately identifying down devices / sites. Normally we would have an edge router, but since we recently pulled this other network entity into our Spectrum fold. and it is all on its own landscape, there are no edge routers with a clear point to point connection on this network. Just a large number of L3 switches trunked to each other on the same subnet / vlan.
Looking for suggestions.
Thanks ...