No, I was thinking that I-NCO [from AgNCO + I2 -> INCO + AgI(s)] would be added across the double bond of safrole, making the iodine sticking off the terminal carbon, and the -NCO sticking out from the 2-position. LAH reduction would dehalogenate the alkyl iodide to the 2-isocyanatosafrole, and at the same time reducing the isocyanate group to H2O and -NHCH3.
My idea was not to make the oxazolidine, just perform the iodoisocyanation and then reduce.
Also, in regards to your question about the N/O-orientation, G. I. Poos says in US Pat 3278382 that 2-amino-4-phenyl-oxazolidines does not have the same central activity as 2-amino-5-phenyl-oxazolidines (aminorex), so I guess it is very important that the molecule stays a phenylethylamine (with the other N/O orientation, it would be a benzylamine).