The main problems with starting on sample roasting are down to differences in why you're roasting which impact you roasting needs to be approached and (to an increasingly lesser extent for new installations) substantially worse instrumentation on smaller machines. It means you end up with no experience dealing with 2nd crack at all and learn a lot less on earlier parts of the roast since you're really trying to do everything the same, often without the tools that make this easier.
While starting with sample roasting does mean that messed up batches waste less coffee, those sample roasts should be getting used in service of high impact purchasing decisions, so mistakes can be a lot more costly unless you're just drilling on things you don't intend use (in which case 100% of the coffee is waste instead of being able to sell the successful batches).
