gLabels is unable to handle CSV files encoded as 16 or 32 bit Unicode characters as a merge source. No errors are given, but the data is mostly meaningless. One source of such files is contact information exported from GMail in "Google CSV" format. (This, apparently, data-specific as I find it being used only for certain selections of contacts in my file).
When not the system default, 16 or 32 bit Unicode formats are hard to detectg automatically unless they include a "byte order mark", in which case they can reliably be detected by examing the first four bytes of the file. GMail flags its files as UTF-16LE when exporting as such.
I have attached a sample UTF-16LE encoded file with a BOM as an example. It's not a Google Contact file but includes a wide sampling of non-ASCII characters.
See also Feature Request #110 Support UTF-16 CSV files as a merge source (same issue last year).