A net that connects to a harness entry gets implicitely named HarnessName.EntryName, so for instance the net that connects to pin 1 on P1 would wind up named Input.1. But then in your subsheet, you explicitly name that net Input1, hence the multiple names. In the Project Options you'll find a few different options for net naming schemes that you can play with to affect how this gets resolved.
In the end, whether or not you permit multiple net names is up to you; it shouldn't cause any problems other than possible confusion when you go to lay out the PCB and the net name you see there may not be the one that you expect. If you look at the schematics for ST's Discovery boards, for example, they have up to three or four names on each net. It's ugly and annoying to read, but it does work.
If I were you, I would probably try defining the differential pair names in the harness entries, so instead of 1,2,3..., name them things like DP1_N, DP1_P, DP2_N, etc. This also gives more clear indication of the functions of the nets within the harness. (That's the whole point of using Harnesses instead of Buses anyway!) Then you should be able to leave off the explicit net names you have in the subsheet and let the port names define the harness names. That should give you the suffixes you need without any redundant net names.