Did you run any performance tests with this? It LOOKS simpler to the eye and at first I thought it looked really good--and it still may be a better alternative than the Itzik Ben-Gan algorithm. I ran some quickie tests with both by plugging your new code into a tally table itvf I use frequently which uses the Itzik method. Then I ran a very simple function that stepped through random characters in strings of varying length. Just something to make a comparison.
I plan on doing some more testing but it will probably be next week if I'm to do it right with some good functions. If you have any ideas for some functions/scripts that might give your model a good test, please pass them along. I've been working on creating a testing methodology that will (I hope) be easy enough to use so that any promising ideas like yours can be put to the test.
Now things might change with different applications, but the bad news at the moment is that old Itzik still seems to have a slight advantage. I've attached a screenshot of the intial test runs.