Good to hear.
Since the spectral cleaning algorythm looks at the data on either side of your selection (in it's typical setting) in doing the reconstruction, the object that contains that selection probably has to be somewhat longer in time than the selection itself. How much more I have no idea.
Makes sense. I'll go back and lengthen my object even more so I have additional
data each side of my selection and see if that makes any difference in this case.