Most pipeline structures fetches register operand at Instruction Decode stage as far as i knew. Without operand forwarding, i don’t know how they are getting F2 value for the 3rd Instruction from 2nd instruction before the WB stage of 2nd instruction.
Although, there could be multiple structures for a given pipeline. All, would lead to 19 cycles if packed properly.