One thing you could try is to vary the PDF input scaling. Increasing the scaling essentially injects more white space in between the fields. This might help Monarch fit in the "hidden" text without making the rest of the line vary from the rest of the report.
A second approach would be to keep the scaling, keep the trap characters as they are, but to highlight the whole line as a field (call it Blob), and then use calculated fields to chop the line into sections.
If the Transaction Type field is the only one that varies, then the fields reading across would be extracted from Blob with:
Intrim() removes extra spaces, lsplit and rsplit chop the line up into sections based on each " ", extract() selects the area between one string and another.