G'day Troy,
Yes survey "Class" is verified by analysis of a semi-major relative error ellipses based on a minimally constrained adjustment but at this point has no connection to datum so is an indication of internal fit only.
For this reason this term "standard of accuracy" used in relation to "Class" is a little misleading as "Class" is an internal relationship (determined from network design, survey methods, equipment, techniques etc) of the strength of the network but without connection to datum then accuracy of the station coordinates are not known at this point.
Order on the other hand is determined by a similar error ellipse analysis but based on connection to a known datum.
For example if our required "Class B" survey was then connected to Order 0 (Class 2A) marks then the best Class/Order that can be achieved is Class B, Order 2 eventhough the "accuracy" of the resulting coordinates are probably better than second order, the controlling factor in this example is the Class, being B
Connect this same Class B network to Order 5 control (or worse) then the best we can claim is Order 5 (or worse) even though the survey is to Class B the accuracy of the control is relative to the controlling marks. A similar scenario would also occur if the fully constrained adjustment error ellipse result indicated a poor quality Order, still Class B but poor quality coords yet meets the requirements of the legislation being Class B.
In both examples we still have the same survey Class but the resulting accuracy of the coordinates would be quite different and I believe this is why the legislation is not adequate in this regard, Class by itself does not control accuracy.