SQL & PL/SQL

I want to use the INSERT ALL INTO style of multiple-inserts (with 10gR2) to normalize a wide table into a skinny table. The large table has a few fk columns and 100 numeric columns. I want to insert them into another table as 100 rows in a single SQL pass.

CREATE TABLE WIDE_SOURCE(
FKEY_1 INTEGER,
FKEY_2 INTEGER,
FKEY_3 INTEGER,
AMT001 INTEGER,
AMT002 INTEGER,
...
AMT100 INTEGER);

CREATE TABLE SKINNY_DEST(
PKEY INTEGER NOT NULL,
FKEY_1 INTEGER,
FKEY_2 INTEGER,
FKEY_3 INTEGER,
AMT INTEGER,
CONSTRAINT PK_SKINNY_DEST PRIMARY KEY(PKEY)
);

CREATE SEQUENCE KEY_GEN MINVALUE 1 INCREMENT BY 100 START WITH 1;

INSERT ALL
INTO SKINNY_DEST VALUES(KEY_GEN.NEXTVAL+001, FKEY_1, FKEY_2, FKEY_3, AMT001)
INTO SKINNY_DEST VALUES(KEY_GEN.NEXTVAL+002, FKEY_1, FKEY_2, FKEY_3, AMT002)
INTO SKINNY_DEST VALUES(KEY_GEN.NEXTVAL+003, FKEY_1, FKEY_2, FKEY_3, AMT003)
...
INTO SKINNY_DEST VALUES(KEY_GEN.NEXTVAL+100, FKEY_1, FKEY_2, FKEY_3, AMT100)
SELECT * FROM WIDE_SOURCE;

With this query, each row in WIDE_SOURCE will generate 100 rows in SKINNY_DEST. I know that works. Unfortunately, it appears that the sequence isn't working the way I thought it should- with multiple nodes and parallel insert, there's some overlap in the generated key.

Is there a better way? I've considered 100 separate sequences, but that just seems plain silly. Am I trying to have my cake (the parallelism) and eat it too (by disregarding limitations of Oracle's sequence construct)?