Rubbersheet process changed from ArcMap to ArcPro

[BillyBuerger]

I recently started doing some rubbersheeting in ArcGIS Pro.  I've done a lot of it in ArcMap over the years and have some automated processes in place for creating and processing the data that needs spatial adjustment.  But I noticed I was getting some unexpected results in Pro compared to what I was used to in ArcMap.  For a very basic example, in ArcMap, if I place a single displacement link between two known source/target points and run the adjustment, all of the features affected by the adjustment will move per that single link.  In this example, I overlaid the original data in red with the adjusted data in grey with one displacement link in on the right.  You can see how that same shift is applied to all features.  Where the data is off by that same shift.  There is no skewing of the data, just a shift...

But, doing this same test in Pro gives a very different result.  The shift at the link matches as expected.  But everything away from that link starts skewing.  It seems to want to assume there are anchors at the edge of the area being adjusted so that the farther you are from the link and closer to the edge, the less impact the link has on the shift.  This can be seen as the parcels on the right side shifted to the right as expected but the parcels on the left only moved to the right ever so slightly.  Same with the top/bottom.  The parcel edge along the top didn't shift down nearly as much as the right-of-way lines near the link...

This is a very simple example and when really adjusting this data, I would include a larger number of links.  But the same issue applies still.  If I create links near the right of ways but not the back sides of the parcels, the right of ways adjust as expected but the back sides of the parcel do not because it's near the edge and is affected by the apparent anchor along the outside of the adjustment area.  Meaning I need more links that I would otherwise expect to try to cancel the affect of this apparent boundary anchor.

Is there any documentation about what changes were made in the rubbersheeting algorithms used in Pro compared to ArcMap.  There aren't many additional options in the transform tool so I don't assume there's any way to make Pro act more like Map.  But if there is something I'm missing, I would love to know what that might be.  I can see how in some cases, this change may be desired.  But in other cases, including mine, this behavior breaks my processes and will require some major rework to make it work as intended in Pro.

[DanPatterson]

I can't remember back to arcmap, but if memory serves the process is somewhat different being more emphatic about creating the links first then rubbersheeting rather than the process you describe

Generate Rubbersheet Links (Editing)—ArcGIS Pro | Documentation  first, then

Rubbersheet Features (Editing)—ArcGIS Pro | Documentation

 if this is what you are describing

[BillyBuerger]

Yes, I'm generating links and then applying the transformation.  I'm testing using the editing tools which is what I was describing in my post.  I did try using the geoprocessing tools as well to see if those gave any different results and got the same.  Except I didn't use the generate links tool and manually placed the links.  In the end, my process involves using the Rubbersheet method from the EditOperation API in an add-in tool.  But again, this is all the same processes, just different ways to execute them.  It's the rubbersheet logic that I'm asking about as that seems to have changed in Pro compared to ArcMap.

[BillyBuerger · Accepted solution]

I found a work around for my issue.  Since the rubbersheet treats the boundary (implied if there is no boundary based on the input features) as an anchor that doesn't move, you can specify a boundary that is much larger then the features you are rubbersheeting.  The skewing will still happen but since the anchor points are so far away, the amount is insignificant.  In my process, I'm generating point features at the locations I want to rubbersheet so I can just add four more points based on a 100,000' extent around my features, do the rubbersheet and then just ignore those points.  Testing is working as expected.