This is a really interesting question, especially in the context of today where the quest is usually for more detail, higher resolution, etc. To directly answer your question, I think you are performing the exact correct operation.
As I see it, the reason for generalizing a layer is to reduce the size and complexity, for performance reasons. This might be desired when you know that a layer is not going to be viewed at a very small scale, in your case, 1:10,000. What would ultimately be desired, in my opinion, would be to decrease the complexity of the polygons as much as possible, without losing the overall shape of what they are describing.
To get to the point: What you are able to see after using the 50 Meter max offset, is not necessarily as important as what is happening behind the scenes.
You need to look at the number of vertices that are in the polygons before and after you run the simplify routine.
You could create two new fields in your attribute table, "OldCount", and "NewCount". Run this field calculator script "Count Vertices in line or polygon features in ArcMap 10" on the "OldCount" field, then run the "Simplify Features" tool, then run the script again on the "NewCount" field. This will give you a good idea of how well your 50 Meter number worked to generalize your polygons.
You might be surprised to see that the number of vertices is drastically reduced even though you don't see much noticeable degradation of the visual quality of the polygons.
Hope this helps!