When adding definitions to a step in your project you must be aware of the audience you are targeting. Definitions that are Not Applicable to the audience will not be evaluated when that step is processed. If you have definitions for that audience, they must be applicable to at least one target in the sense of OS version and affected products (if product is applicable). The target doesn’t have to vulnerable but it must be applicable.
Thanks for getting back to me. I still don't quite understand or my logic is flawed.
From the looks of it, Sample size is based on: Devices in the Group/Query/Scope/Task MINUS any OS platforms that don't match. (referring to the Sample size hover note)
In this example, we have a scope of devices used for testing of 218 devices. 173 devices are Windows 7 Devices, if we assume that the rollout project doesn't take into account "Affected Products" of the definitions then its correct.
When it comes to Windows 10. There are at least 4 Products,
Windows 10 RTM (Release)
Windows 10 Version 1511 (First Major Update)
Windows 10 Version 1607 (Anniversary Update)
Windows 10 Version 1703 (Creators Update)
If the rollout project only considers OS platforms as the sample size but not the affected product of the definition, then none of the Windows 10 patches will ever reach the 80% success rate due to the rollout looking at all Windows 10 devices.
The only way I can see it working is moving each Windows 10 Product into its own Project Step with a query breaking down Windows 10 devices into its Product version.
Maybe its fixed with a later version.