If the file browser does not refresh, this is a bug. Which operating system do you use? For the loaded metadata you can right click and select refresh (or hit Ctrl+r). Easy way to refresh all loaded releases is hitting Ctrl+a (select all) and Ctrl+r to reload.
Not sure what you mean by mess, can you explain? One thing is what @psychoadept answered, maybe you want to use %albumartist% instead of %artist%? Also if you prefer the featuring info in the track title instead of the artist name there is the “Feat. Artists in Titles” plugin (see Available Plugins - MusicBrainz Picard)
We can’t help you without knowing what tagging problems you face here 
Check Options > File naming > Windows compatibility. When enabled this will replace all characters illegal in Windows filenames with underscores by default. This replacement is only done in the file names, not in the tags. If this is for some reason not sufficient you can use $replace() in scripting to do additional replacement.
Lookup and scan do different things. Lookup uses the existing metadata to do a query on MuzsicBrainz.org. If you have bad or missing metadata this will fail. Scan on the other looks at the audio using acoustic fingerprinting with AcoustId. It can help on cases where the metadata is not sufficient. The general advise is to:
- First cluster your releases
- Do a lookup on the clusters
- Scan files not matched.
Clustering your unmatched files tries to group your files by album (again based on existing metadata). If you lookup an entire cluster, Picard does a lookup for a matching release, including the track count. This helps in keeping files together. Even if it cannot match one or more of the files in the cluster to individual tracks, it will still put the files to the release where you can then finalize it. Just doing a lookup on unclustered files does a lookup per file. This can easily result in Picard matching the files to correct songs, but on different releases.
Scan on the other hand always works on a per file basis. And since it does not use metadata you are more likely to end up with compilations and files scattered across multiple releases. Still it is very useful to have in difficult cases.