>> 0.98 * 7.61 ------> 1 * 8 (not sure about the 0.98 going to 1 on this one)

>The fourth example, once again, is correct: rounding to 1 s.d. means, in this case, rounding to the closest 0.1; 0.98 is closer to 1.0 than to 0.9. Or just look at the algorithm in the previos paragraph (the 8 is dropped, 0.9 is increased to 1.0).

There is a "fudge" factor inherent in rounding: how big is the difference percentage-wise to the rounded operand.
In case of the fourth example I'd be perfectly happy to spare myself the effort to round up to 8 and leave it at 7.61, which is much closer to the actual result. Even true for 1.02 * 7.61, were at least the sides of the cut off parts are eaqulizing and not both skewing in the same direction.