I can confirm that DCMTK with ICU enabled does not support the same set of DICOM character sets (i.e. all) in contrast to the original implementation based on libiconv. That's the reason why I personally would still prefer libiconv over the other two character set options in DCMTK.
By the way, this is a well-known issue for the character set conversion support of DCMTK. I hope that the author of the ICU and stdlibc (iconv) support will fix it anytime soon.
