Now this is something to give a read up on. If I had my pc running it would be a lot easier to give examples, but I guess all I can do now is refute your arguments.
When you say that each bar is taken as a unit, what you really mean to say is that each phrase is taken as a unit. What this means is that the phrase now becomes the independent variable, while the number of measures depend on the length of the phrase. Using this method cannot tell you whether the song is 180 or 90 bpm, so it is dismissed.
Your point on 'regular beats' is a little flawed since you deny the possibility of 0 or 2 or more regular (irregular) beats. If the song is 'mainly' 1/4, for example, it certainly does not have to be introduced by 1/4. Similarly, you dismiss the possible existence of a second beat syncopation, when really, one can easily just compose a piece with that said syncopation and render your point invalid.
What's more, it appears that my point raised about the bass>snare>bass>snare defining the measure has not been addressed, while it easily remains the simplest way of finding bpm. I, too, had my moments when I was bound by the strict rules of classical music, but this should evidently not be the case, don't you agree? :3
Note, this is not about me trying to prove that I'm correct; rather, it is more about getting the most things correct on a map that you can actually control before I (we) rank it.