Subtract sub-groups not reliably working

[jbatterman]

Describe the bug
Subtracting sub-group members from a group does not always remove them from the group on evaluation or during computations.

To Reproduce
Steps to reproduce the behavior:

1. On NWO T7 with NWDM database, open Time Series Group Editor.
2. Open Stage-Orifice/Radar, Raw-LRGS-NWDM, and Raw-WebPull groups.
3. In the Stage-Orifice/Radar group, click Evaluate. Note that the group contains ACIA.Stage-Orifice.Inst.15Minutes.0.Raw-LRGS.
4. In the Raw-WebPull group, click Evaluate. Note that the group contains BIL.Stage.Inst.15Minutes.0.Raw-USGS-Web.
5. In the Raw-LRGS-NWDM group, note that Raw-WebPull and Stage-Orifice/Radar are both listed as Sub-Group Members with the "Subtract" combination.
6. Click Evaluate in the Raw-LRGS-NWDM group. Note that ACIA.Stage-Orifice.Inst.15Minutes.0.Raw-LRGS evaluates to be in the group, while BIL.Stage.Inst.15Minutes.0.Raw-USGS-Web does not.

Expected behavior
Expect both time series to be subtracted from the Raw-LRGS-NWDM group.

Screenshots
Stage-Orifice/Radar group containing ACIA.Stage-Orifice.Inst.15Minutes.0.Raw-LRGS:

Raw-WebPull group containing BIL.Stage.Inst.15Minutes.0.Raw-USGS-Web:

Raw-LRGS-NWDM group with Raw-WebPull and Stage-Orifice/Radar groups subtracted:

Raw-LRGS-NWDM group containing ACIA.Stage-Orifice.Inst.15Minutes.0.Raw-LRGS:

Raw-LRGS-NWDM group not containing BIL.Stage.Inst.15Minutes.0.Raw-USGS-Web:

Desktop/ServerSide (please complete the following information):

• OS: Windows 10, Client 7.0.1
• Version: Server 6.8

Additional context
For some reason the compimport did not import these groups properly to the HEC T7 and we are missing key data in the subgroup mentioned for this bug. The Stage-Orifice/Radar group is blank on the HEC T7.

[agilmore2]

I've also seen this. I think Mike Maloney explained some of this as the subtraction working at the timeseries level, but the timeseries expansions to cover variable site, constant parameter occur after or some such. This definitely could use some work.

[MikeNeilson]

I suspect, and I haven't dug in yet - hopefully someone else will, it's an order of operations issue. It doesn't appear you can control when the subtractions are done.

It's logical that everything gets expanded to a time series, combined, and then the subtractions removed, but that's just me. There's a whole lot of set theory this sub-system gets into.