Parsing using appendValueReduced relies on the effective chronology to convert the value.
But if the formatter consists of something like

appendReduced(...).appendChronology(...)

then the chronology won't be correct when the reduced calculation is
performed (as it is parsed after the reduced field).

What is needed is a new bit of code at the end of the parsing phase
which tidies up incomplete reduced fields. There are a variety of
possible designs and it is all internal.