One of my favorite experiences was playing with a pickup partner in a midnight swiss who played a home-spun strong club system. His base structure was odd. I agreed to play his system (why not?) and asked for the meanings of the opening bids. When he started to tell me the responses and rebids, I cut him off and said, "Let's play -- that works." We ended up winning the midnight swiss, with almost no bidding misunderstandings. How?
Bridge theory is fairly predictable, I think. There are a set of general principles that will emerge from context, often very complicated but nonetheless present. These can be predicted.
Granted, some situations call for choice. For instance, consider a Mini-Roman 2D opening. There will be a shortness ask, and that will usually be 2NT. The shortness will be told in steps, either bidding ther stiff or bidding one under the stiff, usually.
Within the context of any start, therefore, there is a predictable set of defaults. Partnership agreement seems to be a combination of tasks, therefore. First, we start with definition of various starts. A "start" could be mid-auction, but nonetheless it is a "start." The start will usually call for a logical set of follow-ups, based on general plausibility defaults. However, there may be two or more plausible defaults, which call for meta-agreements. Some plausibles are "better" than others, and sometimes understanding the reasons for calls provides and answer to the PD's. At times, however, discussion yields a more artificial "improvement" to the PD. These are discussed and perhaps memorized.
Good structure involves a combination of these factors intertwined in as consistent a manner as possible. Thus, if shortness asks tend to be in steps, they should probably be in steps always. If shortness bids are "one under," this should be a repeated theme. Repetition of theme and defaults enables understanding in undiscussed parallels.