Much worse? or more in line with what most readers want: simplicity?
The 3 significant digits is the raw full-precision data, if you want to do more with the graph than get a brief impression. This in contrast to having to infer an approximation by following lines to a coarse axis and visually interpolating the desired values. Providing this data obviates objections #2 & #3.
Don't let the tickmarks, there or not, distract you from the zero. A favored lie in statistics is to conceal the lack (or off-the-chart) of zero, betting that the reader won't notice it having been dazzled by the dramatic angles on what (in many cases) is a rather flat data set. Default assumption should be the origin IS zero; if you're hiding the zero, ask yourself if you're really trying to lie to the reader. Marking the origin may provide plausible deniability, but most readers won't notice it - so act on truth, rather than concealing the lie in plain sight.
Comparison should be as natural as the context & form of the graph makes appropriate. If it's not suitably obvious, reconsider the graph type. If the reader wants more accuracy, a graph may not be the best medium - and golly there's the raw data.
The context should make the axis labels obvious and redundant. If it's not already obvious, reconsider whether the title or other commentary can be reworded to make it obvious.
Of course, all this depends on what you're really trying to tell the reader, and what form follows that function. If tick marks or labels or origin value is necessary, include it ... but whatever isn't necessary, drop it.