But the strange thing is why the micros are different in behavior ??? Does ATMEL produce micros with different noise immunity ??
This is not Atmel, but Philips, I am sure Atmel is similar in this respect.

HIGH level input voltage min 3.15 typ 2.4 V
LOW level input voltage typ 2.1 max 1.35 V

As you see there is an allowable tolerance for the point where a CMOS input triggers which, I believe, is the reason you see the difference between the chips. E. g. if a high input spikes down to say 1.8 V some devices will respond, others not, in accordance with the tolerances specified in the datasheet.

This again proves that Testing can not show the absence of bugs only the absence of known bugs.

Erik