Actually, if you read Diffie and Hellman's paper closely, you'll see that they explicitly talk about taking another's party value from a public file. Thus, it really already does public-key encryption. However, they didn't call it that, and people didn't view it as public-key encryption, but rather as key distribution.
The reason for this is that at the time, the understanding was that public-key encryption required a "trapdoor function" that can be inverted using the secret key. In fact, this is how Diffie and Hellman defined it, and thus they couldn't achieve it.
Now, El Gamal explicitly stated that the technical construction is just Diffie-Hellman. However, his conceptual contribution is enormous; he recognized that public-key encryption does not need to be an invertible function. By the way, the majority of his paper is about the digital signatures and this is novel.
This discussion raises an interesting point which is that some important results are completely non-technical and are important because they make us look at things differently. This is exactly what the El Gamal paper did.