Bugreport explains it:
This is a special case of the general problem, "show me errors, in version Y, of types that were also captured in version X".
For a particularly thorny example, imagine that Ubuntu 12.10 reports hangs of 30 seconds or longer, while Ubuntu 13.04 ratchets that down to 10 seconds or longer. On the graph, hangs of over 30 seconds should show up on the "13.04 by 12.10 standards" line, while hangs of between 10 and 30 seconds should not.
and
This is fixed in production. We now have a separate line for "by 12.04 standards." Matthew, you mentioned that we need some way of tracking the usage of recoverable errors, as we don't want its usage growth among developers to add noise to the graph. I've filed bug 1084079 for that.
There is also bugreport that has an explanation:
Error reports collected from Ubuntu 12.04 include application crashes and thread crashes.
Error reports collected from Ubuntu 12.10 include application crashes, thread crashes, and recoverable application errors.
This may account for much of the current difference between Ubuntu 12.04 and 12.10 in the http://errors.ubuntu.com/ graph, since the two most common errors currently are recoverable application errors in Ubuntu Software Center.
This can be fixed by Ubuntu 12.10 having a "by 12.04 standards" dotted line on the graph.