It could be the way it handles communications, I have seen various problems in the past, recently when I was developing MB protocol on an FX with standard 232 adapter module I noticed that on the MB simulator there was originally an error every other request, cured this by a small delay between actual recieve from the salve & the next request although there was an inherrant delay of 40ms due to scan time between a recieve of the reply before sending the next request, this was due to the send request being at the start of the program so had to process the plc program to the end before the recieved data was processed & a reset of the recieve bit from the 232 card.
Even when I got it to work without any errors I noticed that the simulator RX/TX data was showing part or all of the CRC data twice once in the actual request & again afterwards, it was almost as if my plc was sending the correct message but then sending the last one or two bytes again, however, another test using a data capture program confirmed it was not.
It did fox me at first as the simulator was not showing any bad transactions as i would expect if the datagram was incorrect, again tested that it was not the simulator I sent messages for example a wrong CRC, missing CRC & missing bytes from the data, everytime the simulator picked this up as a bad format. the only thing I can think of was the simulator had a bug where in com monitor mode it was populating the recieved message 7 then for some reason duplicating the last couple of bytes.
It is known that the modbus protocol has many variations probably not in the original design but interpreted by others with slight tweaks i.e. byte swapping 0 or 1 base addresses using the 4000 - 64000 range rather than a base offset of 0/1. It does get confusing.