Well, with NO delays, it's going to try to send/recieve as fast as your server can give responses...so you might have some non-sychronization. I add my own delays (since recorded delays might not be what some people really want in a test), and never have "none", since that's not representing what a user could do.
And TTLB is BAD if you don't wrap it so it's not checked when you might time out, or get an error (like a 404). If you check TTLB in those situations, the script will crash. You might want to consider using a variable to measure change in the Timer function, and wrap your request.send with that. That way you'll have time whether you time out, get a non 200 http response, or whatever.
And the CSV is something that I do presently, it's not 100% accurate (mostly b/c of what i said in the first paragraph), but it does give you an approximate way to measure performance. Which can be very useful if you're making performance CHANGES...so you can do a before/after. If you're looking to measure for someone how many users can do "x" on this site at a given time...the best you can do with ACT (in my opinion) is make an educated guess. Even the most complicated algoritm that I've made in ACT to simulate many different user actions and activities and doing so in a repeatable manner, still doesn't reproduce exactly what all of our users do at a given time. So I use it mainly to test performance enhancements and system changes...because it's more easily quantifiable.