Should I update existing administrative boundaries to reflect latest OS data?

I recently added ward boundaries for the Manchester City Council district area to OSM, which were derived from OS Boundary-Line data (latest update October 2018).
I checked that the outer perimeter of all the ward boundaries is coincident with the overall district boundary for Manchester in the same dataset, and it is. However this current Ordnance Survey boundary seems to have numerous (small) discrepancies from the current OSM administrative (level 8) city boundary (relation 146656). This could be because:

• The current OSM boundary is from a different source than the OS (in which case, which has precedence?)
• The boundary has been legally changed since the OSM boundary was uploaded
• Various boundary nodes have at some point been moved (accidentally or otherwise)

So does the latest OS data for such boundaries take precedence over existing OSM data? Would I be justified in creating new admin boundary relation(s) from the OS data, copying the tags, and deleting the current boundary relations? Alternatively if I leave everything as is, then according to the current OSM model there are small slivers of land dotted around Manchester which are apparently in Manchester but not in a Manchester ward, or vice versa - which seems rather unsatisfactory.