PageStream incorrectly handles encoding of imported PDFs - of both generated by PageStream itself and from external sources.
Left: initial document, which is later saved as PDF. Right: the resulting PDF opened by PageStream.

I assume this is only for characters which are outside of the typical latin-1 code range ? If you can provide a small example to reproduce the issue, it might be easier to fix.

I prepared three single word PDFs. They are generated by different programs, namely by PGS itself, MS Word and Libre Office Writer.
https://0x0.st/HNXJ.lha

Ok, thanks, this clearly shows that all programs use different ways to map the character space inside the exported Pdf, also that both MS Word and Libre Office Writer only export a subset of the used font if you look at its file sizes, with PGS most likely exporting all of the fontdata, resulting in a 100K+ file.
While they all seem to show OK in VPdf, the reimport into PageStream fails to map the "extended" characters 'Č' and 'Š' back to its original mapping again.