I appreciate your input, and I generally agree my representation is not optimal for extension, but I want to clarify what my algorithm does, because this is not a correct understanding of the representation.
The algorithm is based on complement arithmetic, which is why I didn't give it too much thought. The Japanese soroban is based upon it; ie. When subtracting 2 - 8, you add the complement of 8, in base 10 is 2, because 2+8 = 10. The problem of 2-8 becomes (2+2)-10 = -6. In base 60, we would add the complement of 60 and subtract a 1 from the column on the left, until we get to the degrees, which are in decimal format.
For the very reasons you outlined (Perl's representation of small negatives), I represent -0.1 as an equivalent polynomial (-1 + 9/10),
or in this case 0 -1 0 as (-1 + 59/60 + 0), with the denominator implied as (-1 59 0).
Algebraically, a + (-a) = 0. If (0 -1 0) + (0 1 0) = (0 0 0), then we must treat -1 59 0 as equivalent to 0 -1 0, because
(-1 59 0) + (0 1 0) = (-1 60 0) = (0 0 0).