Knowledge Based Teaching - a small challenge

I am not just trying to be provocative to those who strongly believe in some version of Direct Instruction - teach knowledge, no fluffy stuff poorly paraphrases what they believe. well I am, in the hope that they will respond and show me where I am getting it wrong. Or, perhaps, let them carry on with their own thinking and delve deeper into the learning process.

One of the tenets of Direct Instruction is to present the knowledge that we want students to learn as clearly as possible. I asked @oldandreuk if he would ever make learning more difficult for students. He answered in a slightly reserved fashion saying that he probably would not.

I later read some research that seems counter intuitive. The research found that if students were presented with some material to be learned they learnt MORE and RETAINED MORE if a less clear font was used in the text than if a standard, clearer font was used.

Ok, from that I make a leap, untested, but playing with ideas is fun, to hypothesise why it might be that the less clear font, which makes students work harder might be working. What I came up with depends somewhat on accepting learning is a process of either creating a new memory map in our brain or, more usually, adding to an existing memory map. If we make students work harder, make the learning process more challenging without making the content to be learned more difficult we force them into activating more brain memory maps, more schema as the cognitive scientists say. More links mean better learning as it creates stronger memories as more maps could be activated to retrieve the learning and attaching to more maps means a deeper understanding.

May be I have created a load of nonsense but it could be correct.

Why am I exploring this?

Because although I agree with lots of the Direct Instruction stuff it makes me uncomfortable. It seems too barren a way of learning. Too programmed. Also I think DI works better for some types of learning than for others. Maths, MFL, some science work are examples of linear subjects where knowledge is built from previous learning. For example, in maths if you do not understand how to multiply by 10 you will certainly not understand how to multiply by 100.  Times 100 builds on the understanding of times 10. It also does not make for the best maths teaching to learn times 100 before learning times 10. But aspects of, say, English are not linear - some English learning will be but the great proportion of maths is linear and English is less so.

Also, research shows that DI is about 10% better than other ways of being taught. Not an efficiency improvement to be sniffed at but if other methods are so poor and clearly flawed, according to DI folk, then why are other methods getting close? Why do they get to within 90% of the claimed DI?

Be glad to hear your disagreements.