In my Open Roads testing, I have created a Civil XIN dgnlib and a Survey XIN dgnlib. It was necessary to set all of the survey settings in both of these to the appropriate settings, even though my active file was already set correctly. This may prove problematical as one may need to change some settings for a different type of survey file. But once that was set, all of my cells, coming from a special library with a mixture of annotative and non-annotative cells, were coming in at 12X the scale that they should. And when I tried changing the scale factor, while my non-physical linestyles changed correctly, my cells did not.
Our local DOT workspace, which is my test workspace, uses a somewhat unique scales.def and sheetsizes.def files. Their sheet sizes are listed as units of Survey Feet, where the height and width numbers are actually the sizes for inches. So an 22 X 34 sheet is listed as 22 X 34 but the units are feet. So in the scales.def file, all of the imperial scales are defined the same as ratio scale - 1" = 50' and 1:50 are identical. On the default files from Bentley, 1"=50' is the same as 1:600. So its possible that the scales are being effected by this, but I am not sure.
You can test this using the MD SHA workspace and XIN file. Keep in mind, their cell library does not have annotative cells - I had to modify a copy of their PSD_TOPO.cel library to turn most cell in it annotative, except for drainage inlets and pavement marking words and symbols. Also, many of their survey features are not defined as surface features, so Open Roads draws most on level Default.