Yes, sorry, I really meant to say the PC (rather than the modem) up, as it is that which is running the 1min harvesting. My mistake, although if the modem had been down, then it can't be sampled then either.
I'm afraid I don't recognise the version numbers that you refer to, but that's just my faulty recollection. v1.0 was BE's first 'compiled' version of the package, and v1.1 was a more recent update to that. I cannot recall what version number I had previously when it was the .bat files.
So, yes, a combination of switching off the PC, the harvesting issues with v1.1, and the not-immediately-obvious semantics of the Ongoing stats plotting, may well result in it searching a 'sparse' modem_stats.log file for the required number of samples, resulting in the bunching that you are seeing.
I appreciate that this does not fully explain why 'it worked before', but then, it didn't have this harvesting issue then either. I think therefore your conclusion is probably correct.
If you can find & download Textpad, or some other similar editor that can give you record numbering, you should be able to find out
exactly how many real samples you have in your file.
HTH