Robert,
Very interesting stuff, thank you. One factor that would account for at least part of the discrepancies is the OR that you are using - let me rerun the NLs with a 10 min OR and repost the results.
Also, I've just sent you a PM, please take a look.