That's only part of it, the major problem with classic UUIDs are that writes occur randomly across the entire keyspace. Whether it's represented as a string or in binary, ordering does not change, and so the underlying indexing method must cope with these random orderings at insertion time.
At least btrees perform much worse with random insertions. I don't know how much impact it has on LSM
LSMs and fractal trees handle random inserts much better. But you still lose locality of lookups: items generated near each other time-wise are not correlated on disk.
At least btrees perform much worse with random insertions. I don't know how much impact it has on LSM