Why do the floating point registers read invalid or incorrect values?

You need to reconfigure your Modbus Master software to interpret the first 16-bit register of the 32-bit float register pair as the least significant 16-bit word. This is usually done with a selection called “Swapped FP” or “Reversed Float”. If that doesn’t work then try handling it the other way by selecting just “32-bit Float” or “Floating Pt”. If you are writing your own software, see Floating Point Register Details for the Modbus WattNode.