The VCL coordinate system is vertically flipped compared to the Quartz coordinate system. There are 
two possible approaches to handle this:
1. use a y-flipped height-shifted CTM for the graphics context to allow using VCL coordinates directly
2. flip y-coordinates manually before operating on a unity-CTM graphics context

OOo's  aqua-port choose method 1 and both approaches should work the same. Unfortunately it shows 
that OSX implements some improvements only for unity-CTM graphics contexts. Moving the baseline 
of text to pixel boundaries is probably the most important thing that is missing in y-flipped contexts, 
which cannot be emulated by moving the text coordinates. This results in grayish text baselines for y-
flipped contexts whereas a unity-CTM graphics context gets higher contrast baselines.

Unless Quartz/ATSUI/CoreText provide the chance to get the same look also for flipped-CTM graphics 
contexts we should consider to re-implement the VCL->Quartz glue code with approach 2 mentioned 
above.