This turned out to be a very interesting bug. The file in question showed a couple issue with our cmap parsers for both the Pro and OS version of ICEpdf.
Both the pro and os version of the cmap did not correct handle the cmap entries with values for beginbfrange in the format <src1> <srcn> [<dest1> <dest2> ...]
Also the OS version did not correctly handle multiple entires for beginbfchar and beginbfrange definitions.
As a result both the OS and PRO version do a much better job at text extraction and font substitution. I've increased the severity of the cmap parsing errors so that they will be more visible when they occur, hopefully this will help identify any future issue.
CID test file