All of us have been making 89x5x work with serial ports at standard baudrates.

Recently I had to make one board work with an odd baudrate - with serial port parameters 5787,E,8,1, to communicate with a pump controller. Solution was to go for 5760 (attainable with 11.0592MHz crystal) and it worked alright since error was less than 2%. Processor was able to communicate with the pump controller.

Then I realised that 5787 can be achieved with a 10MHz crystal with Timer 1 constant=247, SMOD bit=1

I measured the bit width to be in the range of 168-170us ( I don't have a storage oscilloscope) . But with this, the processor could not communicate with the pump controller.

Only crystal changed from 11.0592 to 10MHz. No other change in the board.
Your views on this please.

Only Timer1 constant changed from 251 to 247 and SMOD bit changed from 0 to 1. No other change in software.

Can somebody think of an explanation for this? Pump controller document says it is working at 5787 baud. COuld it be working at some other baudrate? Will somebody purposely give wrong doumentation like this? Have any of you experienced this kind of purposeful obfuscating of information? Why would one select such odd baudrate? Is there any standard which needs 5787 baudrate?