kdbg: frames resolved incorrectly at backtrace

Created an attachment (id=6559)
backtrace from CMake

While retesting bug http://www.reactos.org/bugzilla/show_bug.cgi?id=5931 on revision 52667 CMake, VBox 4.0.10

1. Boot to third stage;
2. Enabling catching first exceptions via command: set condition * first always;
3. cont and type ipconfig /all in cmd to replicate the CORE-5378;
4. When first exception is caught, type bt

The call stack is resolved differently, than one from CORE-5378, which was raddr2lined manually on rbuild.

See http://www.reactos.org/bugzilla/attachment.cgi?id=5796 for rbuild and attached log for CMake. Assigning bug to Art.