This FAQ describe why ambiguity codes and N's can be present in a read mapping, even that none was present in the un-mapped paired-end reads.

The presence of ambiguity codes and N's in mapped reads can both relate to the use of overlapping paired reads for which there is a conflicting base, insertion or deletion between the two reads. When paired reads overlap they will appear as one concatenated blue read, for which conflicts will be shown with an ambiguity code and insertion/deletions with an N.
 

Ambiguity code in mapped reads:

An ambiguity code will be assigned to any base where the two members of a pair of reads overlap, but do not agree on the base call.
 

N's in mapped reads:

An N will be assigned to any position where the two members of a pair of reads overlap, but there is an insert or deletion in one of the reads on the base call.
 

How is this information handled by the Workbench variant callers:

The Fixed Ploidy and Low Frequency Variant Detection tools will ignore overlapping reads that do not agree about the variant base, whereas the Basic Variant Detection tool will consider the position if only one of the reads passes the quality filter.

More general information about how overlapping paired reads are handled can be found on the following manual page:

CLC Genomics Workbench: 
http://resources.qiagenbioinformatics.com/manuals/clcgenomicsworkbench/current/index.php?manual=Detailed_information_about_overlapping_paired_reads.html