Realistically I don't think there's any real difference, so which ever you prefer is fine.
For some PowerShell commands you can find that one way of doing it is a lot more efficient than another, but not in this case. It tends to be more where you for instance have multiple things in a pipeline, and the different order you do them within the pipeline makes a difference to the end result.
For any command you can use Measure-Command
to determine how long a given command took to complete, but when using that in this instance, eg
Measure-Command {(dir C:\Windows | Measure-Object -Property Length -Sum).Sum} | select TotalMilliSeconds
Measure-Command {dir C:\Windows | Measure-Object -Property Length -Sum | % Sum} | select TotalMilliSeconds
Measure-Command {dir C:\Windows | Measure-Object -Property Length -Sum | Select-Object -ExpandProperty Sum} | select TotalMilliSeconds
you find that you essentially get the same result... or more accurately, there is only a small difference, but it's small enough that if you run it multiple times you'll see different results each time, with perhaps the first version being fastest on one test, and then being slowest on the next test.