Everything with PT in front of it came from the google transit feed relational database or were generated from basic VBA code that I wrote. The "WalkNet" component is a shapefile derived from the road centrelines from the state land information authority.
I've processed them so they all snap properly as shapefiles, however when I then export them to a geodatadabase feature dataset so I can build a network dataset, the WalkNet is no longer snapped, it has moved about 0.00006 of a metre relative to all the PT components, which are all still snapped to each other. So as shapefiles, they're snapped, but as feature classes they're not.
I realise this probably has something to do with resolution, but when I change the default resolution from 0.00001 of a metre to be finer, I can't export datasets at all - they process, and then have fields but no features. I've tried using Geowizards to run snapping processes again after exporting to feature classes, but for some reason GeoWizards won't accept a feature class inside my feature dataset as an output, so that's not helping.
Everything is in the same coordinate system, including all the shapefiles, the feature dataset and the map document I am using to export. From comparing the shapefiles to the feature classes I can see that all the PT features move slightly comapred to the shapefiles, but the WalkNet doesn't move relative to the shapefile. What could be causing this?
If anyone could help me out with this I would be very grateful.