...and clearly distinguish between "reading + decompression throughput" and "just raw reading throughput" in output.