In a testing lab with unlimited funds, matching RAM (matching CAS, the controllers and the size perfectly) can benchmark faster than non-matched RAM of equal total size.
In the real world, you get a max of two RAM slots on the MacBook line and most people don't a) wait on their RAM for any significant amount of time ever b) don't throw away good RAM just because it doesn't match another chip they might put inside. (Also, I'm not saying no one waits on RAM, but that the matched vs unmatched situation doesn't cause people to wait)
For MacBook that are now vintage (especially the plastic MacBook with the removable batteries in black and white case) - the VRAM was part of the RAM and having matching chips was actually noticeable when you didn't get it right. Games might even be unplayable vs playable in some cases.
For anything Mac that is Unibody (not with a removable battery / frame is solid Aluminum) - just buy the most RAM you can afford from a vendor that will accept a return within a week if you mess up picking the right CAS latency and it doesn't work when you insert it. I run Mac Mini servers hard with constrained RAM and just buy a pair of 8 GB sticks for each pair of Macs. They come with 2x2GB so the macs have 10 GB of RAM and there is no measurable speed difference between a matched 8 GB and a non-matched 10GB and since the initial 2 GB stick is free I would never go with a pair of 4 GB sticks in that computer.
As a rule of thumb - adding more total but unmatched memory is always faster than adding less total matched memory on aluminum case Apple products released between 2009 and 2014. (And let's exclude the Mac Pro since that can have enough RAM slots and CPU to complicate the generalization I just made)