The larger model is based around how that the extra heat maintained gets distributed around the world and everyone knows this is far less accurate. The people building this are well aware it's highly limited but by trying to balance the errors in both direction they hope to create a good estimate. The problem is when you start tweaking the heat distribution equations it's easy to find errors in every direction because they are greatly simplified so it's easy for someone to introduce systematic errors by only correcting issues in one direction.
Which is why you provide a high level description but when you release your source code you open yourself up to a lot of fruitless debate. Sort of like someone saying based on your methodology the rocks are really 1/1,000,000th a year older which means your science is a joke and less valid than my consistent theory that god created the world 8 weeks ago. But, if you actually build a model from scratch under similar assumptions it's going to have similar outputs and error bars or look really biased.
Still, in the larger context validating someone's data is basically worthless activity. It's far more important to go out and collect new data, build a new model, run a new simulation, and report those results than it is to try and shortcut the process. The data deluge is making it way to easy for people to validate bad science using the same bad data / model as someone else. Fundamental assumptions in statistics break down when you use the same 10 coin flips to substitute for the 10 million coin flips that cost more than your willing to pay for.
PS: The failure to collect significant quantity's of new data is a hallmark of fringe / bad science. If you really think global warming is a joke or you can be more healthy by eating coco puffs then go collect that data and do your analysis and see if someone can poke holes in your process.