My opinion on this seems to be at odds with a number of others and is possibly wrong, but anyway it is important to understand the scope & limitations of the system we work with .....
The argument that circles should be avoided as a geographic object is as spurious as connecting any two points with a straight line and using the line as an object. The only time the line is valid in any coordinate space is if it represents a line-of-sight between two surveyed points (and even then, light bends). As soon as it represents a property boundary for example, it is really valid only at its ends. Similarly with a circle, you could possibly say that at the time it is created, the centre and one point on the circumference were legitimate points. In both cases, they are just representations of reality and obviously have their place. If a circle or any object is drawn and used in its native projection system, it is defined and remains true in that cartesian system. How good it represents reality is another issue. So, my argument is the more object types the merrier, as long as they are all treated as the approximations they really are. And bring on more metadata (and metadata analysis tools), so we have good descriptions of how the data can be used. An occasion when circles (and other objects) may be a problem is when reprojected from the system they were originally defined in. For example, all objects distort as they move away from the central meridian of a UTM system. Because circles are modelled as an ellipse within a fixed orientation rectangle, the rotation that occurs, because of the projection maths, off the central meridian cant be represented. The ability of MapInfo to reproject so readily has some mixed blessings. The answer here is to limit use of mixed projections or be aware of limitations and impact on length, area and object relationships. It'd be nice if there were a summary somewhere of this stuff - may be in Wiki eventually. Whilst playing with this, I did notice that the select and display representations of a circle drawn in say UTM and displayed in geographic projection, are different!? This difference is more obvious, the further from the central meridian you go (say 300 km, with a 100km radius). In spite of the issue with circles noted above, this would seem like a bug (in 6.5). Why else would two apparently cosmetic versions (display/select) of a circle be different. It can be reproduced by clicking from the circumference to the centre. The circle doesnt actually get selected until closer to the center than the circumference. Phil.