What I posted:
_______________
I       ......I
IV      .....V.
V       .....V.
VI      .....VI
IX      ....X..
X       ....X..
XI      ....X.I
XIV     ....XV.
XV      ....XV.
XVI     ....XVI
XIX     ....X.I
X       ....X..
XL      ....X.. WRONG Should be: ...L...
LX      ...LX..
XC      ....X.. WRONG Should be: ..C....
CLXIX   ..CLX.I
CDXLVI  .D..XVI WRONG Should be: .D.L.VI 
MCMXCVI M.C.XVI
MDCLI   MDCL..I
_______________
[download]

Here's more on what I want to do, and why:

             1 2 3 4 5 6 7
Simple Rule: M D C L X V I
[download]

Input String: IV

Character: I
Desired position: 7
Next char: V (Desired position: 6)
I>V
7>6
Discard I; take V (Output: .......)

Character: V
Desired position: 6
Next char: NONE
V<NONE
Keep V (Output: .....V.)

End of Input; Output: .....V. (Agrees with Posted)


---

Input String: VI

Character: V
Desired position: 6
Next char: I (Desired position: 7)
V<I
6<7
Keep V; consider I (Output: .....V.)

Character: I
Desired position: 7
Next char: NONE
I<NONE
Keep I (Output: .....VI)

End of Input; Output: .....VI (Agrees with Posted)


---

 

IX      ....X.. for the same reason IV      .....V.
XI      ....X.I for the same reason VI      .....VI
[download]

---

Input String: XIV

Character: X
Desired position: 5
Next char: I (Desired position: 7)
X<I
5<7
Keep X; consider I (Output: ....X..)

Character: I
Desired position: 7
Next char: V (Desired position: 6)
I>V
7>6
Discard I; consider V (Output: ....X..)

Character: V
Desired position: 6
Next char: NONE
V<NONE
Keep V (Output: ....XV.)

End of Input; Output: ....XV. (Agrees with Posted)


---

 

XIX     ....X.I
[download]

(Special case because we already "saw" X - See MCMXCVI M.C.XVI for a complete discussion.)

CLXIX   ..CLX.I
[download]

for the same reason XIX ....X.I

---

Input String: XL

Character: X
Desired position: 5
Next char: L (Desired position: 4)
X>L
5>4
Discard X; consider L (Output: .......)

Character: L
Desired position: 4
Next char: NONE
L<NONE
Keep L (Output: ...L...)

End of Input; Output: ...L... (DOES NOT Agree with Posted)


---

Input String: LX

Character: L
Desired position: 4
Next char: X (Desired position: 5)
L<X
4<5
Keep L; consider X (Output: ...L...)

Character: X
Desired position: 5
Next char: NONE
X<NONE
Keep X (Output: ...LX..)

End of Input; Output: ...LX.. (Agrees with Posted)


---

 

XC      ..C.... for the same reason XL    ...L... (DOES NOT Agree with
+ Posted)
[download]

---

Input String: CDXLVI

Character: C
Desired position: 3
Next char: D (Desired position: 2)
C>D
3>2
Discard C; consider D (Output: .......)

Character: D
Desired position: 2
Next char: X (Desired position: 5)
D<X
2<5
Keep D; consider X (Output: .D.....)

Character: X
Desired position: 5
Next char: L (Desired position: 4)
X>L
5>4
Discard X; consider L (Output: .D.....)

Character: L
Desired position: 4
Next char: V (Desired position: 6)
L<V
4<6
Keep L; consider V (Output: .D.L...)

Character: V
Desired position: 6
Next char: I (Desired position: 7)
V<I
6<7
Keep V; consider I (Output: .D.L.V.)

Character: I
Desired position: 7
Next char: NONE
I<NONE
Keep I (Output: .D.L.VI)

End of Input; Output: .D.L.VI (DOES NOT Agree with Posted)


---

Input String: MCMXCVI

Character: M
Desired position: 1
Next char: C (Desired position: 3)
M<C
1<3
Keep M; consider C (Output: M......)

Character: C
Desired position: 3
Next char: M (Desired position: 1)
C>M
3>1 BUT... we've already "seen" M...
Keep C; discard second M; consider X (Output: M.C....)

Character: X
Desired position: 5
Next char: C (Desired position: 3)
X>C
5>3 BUT... we've already "seen" C...
Keep X; discard second C; consider V (Output: M.C.X..)

Character: V
Desired position: 6
Next char: I (Desired position: 7)
V<I
6<7
Keep V; consider I (Output: M.C.XV.)

Character: I
Desired position: 7
Next char: NONE
I<NONE
Keep I (Output: M.C.XVI)

End of Input; Output: M.C.XVI (By some miracle, Agrees with Posted)


 

Complex Rule: Discard an earlier, one character match in favor of a longer, later match. In real data, earlier, one character matches also occur and should be discarded, because a longer match is probably on the way. Similarly, we probably want to get rid of characters we've already seen, unless... Unless those characters are SUPPOSED to also occur later, which is why I used the character classes in the regex in the code I posted... Arg.

Desired Output:
_______________
I       ......I
IV      .....V.
V       .....V.
VI      .....VI
IX      ....X..
X       ....X..
XI      ....X.I
XIV     ....XV.
XV      ....XV.
XVI     ....XVI
XIX     ....X.I
X       ....X..
XL      ...L...
LX      ...LX..
XC      ..C....
CLXIX   ..CLX.I 
CDXLVI  .D.L.VI 
MCMXCVI M.C.XVI 
MDCLI   MDCL..I 
_______________
[download]