Good observations.
You might also note that Hong Kong, although technically a "separate" administrative region, is still part of China. The Chinese sent a "medical expert/advisor" to Hong Kong about the time you noted a change in the way Hong Kong was reporting numbers. It is quite possible that Hong Kong was ordered to make their numbers look better.
It should also be noted that Hong Kong had reached the practical limit of their medical system by mid April (duly noted on FR, by the way). They changed their treatment protocol at the same time. Deaths immediately went up, but so did the number of people released. This had the apparent effect of lowering the number of reported SARS victims in the hospital.
Your excellent observation about the change in the way they were classifying patients, combined with a change in treatment that may have been designed to free up beds even if it meant more people would die, may be quite significant.
I continue to hope that the Hong Kong numbers do, in fact, mean what they seem on the surface to mean: Hong Kong is gradually getting SARS under control.
However, if they have messed with their reporting enough, then we may be seeing nothing more than a change in strategy. In that event, the reported numbers do not tell us anything meaningful.
I suspect that they have decided that they can manage a certain number of cases in the hospital system, and that whether or not an early stage victim is immediately hospitalized is decided partly on the basis of bed availability. The good sign here seems to be that the hospitals are now taking more of the early stage patients in, which would indicate a waning of the epidemic.
On another note, have you noticed how closely the China death figures correspond to the Hong Kong death figures? By coincidence, they are issued immediately after the HK figures each day, but unlike the HK figures, give no specifics on who died. Kinda like snatching a number out of thin air.