In the example you gave if you decided to use it you would also need a weighting for the severity of the defect - after all the impact of a word spelled incorrectly against a defect that stopped the application running is significant.
I think, Qualitier, that what you're measuring is effectiveness. The question you're answering is "How effective are you at removing defects via testing?"
Certainly parsing the metrics by severity (with appropriate weighting factors) would be interesting.
Also interesting would be to parse the defects by development phase (e.g. planning > analysis > design > build > test > implementation > operation ...). You'd want to generate a sequence of metrics similar to what you've already presented with M=A'/B' where A'=defects found in the next phase and B'=all defects found since project inception (ex: M(analysis) = A(design) / B(planning + analysis + design).
Clearly, this presumes you have some method of identifying and tracking defects prior to a formal test effort.
There is an article by Ellen George with a approach that tries to predict the number of defect remaining in the system based on the number of defects detected in previous development phases and the defect removal rate usually associated to the quality activies (inspection / tests) of this phase. These values could also be used to allow product to proceed to the next phase. Look to http://www.methodsandtools.com/archi...hive.php?id=13 for a detailled explanation.