Creating multiple non-overlapping buffers
You could set up a model to use the ordinary buffer tool. Buffer the input polygons once. Then buffer the resulting features, and then buffer the next set of resulting features. Then, append two of the result feature classes into the other one.
Seems like it works.
Multiring buffer should include data with the original feature. Unfortunately, as you have found out, that doesn't work if the Donut option is chosen. Here's a relevant Idea: Analysis Tools: Multiple Ring Buffer: Don't dissol... - Esri Community
To Ed's point, it turns out the centroids of each buffer are different, which is super frustrating.
So, here's a toolbox with two tools to fix that.
They turn the disc output of multiring into donuts without sacrificing the attributes.
 | 
|
| Before | After |
One tool (Punch out buffers) works after you've made the buffers, while Clean Donuts buffers the input feature and gets the nice donut result all in one go.
I've only lightly tested these, so your mileage may vary, but it's worked for me so far.
I'll be sharing this toolbox to that Idea as well.
I don't know why either, but here's the proof:
When I was figuring out the code, my initial thought was to go by centroid.
Note that the buffers and the original point are in this coordinate system.
Original point at the bottom, with the other three buffer centroids above.
You also get different stuff if you check arcpy geometry.
So, anyway. They do not have the same centers, which means that if you try to match centroids, you won't get any results.
My solution was to check the ORIG_FID field and work with that. You can check my code to see how I did it.
Full disclosure, I have a hard time with coordinate systems and stuff, so that may be another wrinkle to this, but I tried to keep everything consistent while testing this.
Also, to be fair. That initial test I just showed was buffered geodesic. Buffering planar (again, keeping spatial references consistent) gives a much tighter spread, but a spread regardless.
