Mating methods within in-context assemblies

This relates to top-down, fully parametric models where all parts are dependent on those around them - I have a question/suggestion, but it might help to put it in context first.

I have increasingly been moving away from the common method of having one fixed part, and mating all other parts to that part or each other. By definition, parts created in context of an assembly have their positioin and geometry pre-determined... adding mates afterwards can be problematic (as they need to match exactly the geometry, which can change down the line). This mating method is clearly a necessity for bottom-up assemblies, but seems a bit daft for top down assemblies.

My approach recently has been to mate ALL parts (including the first one) by their reference geometry to the ref geometry of the assembly. Just 3 mates per part, top to top, right to right, front to front. With this method, no matter what changes you make in the future, the parts will ALWAYS be defined, and the mates will NEVER bug.

I'm sure this is a common way of doing things, allthough strangely i havn't been able to find a thread confirming this.

My question is this:

Could I, by the same reasoning, simply 'fix' every part. Although the parts are supposedly fixed in space, they aren't in so far as their positioning is determined by sketches defined by neighbouring parts. ie, they will move according when changes are made to the model.

By doing one of these 2 methods, you are effectively removing the whole 'mate' issue entirely, and thereby massively reducing the error count. It also means that all parts are orientated in the same way. There are a few issues i can think of, but not many. I would be interested to hear people's opinions of this way of doing things. I feel like either i'm missing a very obvious disadvantage, or this is an extremely common approach but isn't documented anywhere!

Thanks in advance

SolidworksAssemblies