Hi Michael,
As a programmer, maybe you can use [link=https://code.google.com/p/mdb-sqlite/]THIS[/link] Java-based tool to convert the TPDB8 file to SQlite3. The page mentions 'SQLite' but [link=http://goobbe.com/questions/2940956/type-4-jdbc-driver-for-ms-access-mdb-database]elsewhere[/link], someone said "I've used jackcess to read the tables from an MDB and convert the database to sqlite3" so I guess SQLite3 is supported too.
> In fact I once for a test converted the TP7 database to a Client-server database (Postgres). With some help from some friendly people here this did work nicely by exporting / importing.
Hm.. that suggests that there are no stray commas or linebreaks in the data.
> I could try to find stray linebreaks but will that really help ?
Maybe you should start by exporting just a few records, and then seeing if the resulting (small) text file can be imported without problems. I'd start with a few simple records, and if that works, try a few that have keywords, annotations and are members of galleries. If that also works, then a larger file should work as well, and I guess any errors that would pop up would be due to stray commas and linebreaks (and maybe quotation marks).
An easier way to find formatting errors is by converting the CSV data to fixed field text. With that, all fields should be neatly organized in columns and all lines should be (nearly) equally long. That makes it easy to spot broken records (too long or short), and if you sort on colums (with a good text editor, I like Boxer), you can locate data that appears in the wrong columns.