I am one of those people who like former D.C. schools chancellor Michelle A. Rhee. The number of Rhee fans has dwindled in the District as she has become more famous and popular across the country. But I still think she did much good for D.C. children — reviving a moribund school headquarters, closing half-empty schools and getting achievement back on an upward path.

So I was unpleasantly surprised to read the statements she made Monday, quoted in newspapers throughout the country, saying people like me were “enemies of school reform” for thinking that a story that day in USA Today had exposed a serious problem of cheating on D.C. tests. The Post quoted her telling talk show host Tavis Smiley that the USA Today story, which revealed more than 10 cases per child in which erasures caused an answer to go from wrong to right in some schools, was not credible and it was wrong to suspect that such changes might have been made by adults throughout the system after the test was over.

Wednesday morning, I got another surprise in the form of a phone call from Rhee. She told me that what she said Monday — her word, repeated often in our conversation — was “stupid.”

She said that she thinks cheating might have occurred in the District and that she is glad her successor, Acting Chancellor Kaya Henderson, ordered a new investigation. Rhee said she still believes that the vast majority of teachers and administrators would never falsify test results, but that there can be exceptions. She said we should improve test security procedures so such abuses could not recur. She said the D.C. schools should ensure that D.C. Comprehensive Assessment System tests, after being completed, are not left in principals’ offices where tampering is possible.

“You have got to have really strong test security protocols at the district level and at the state level,” she said. “The vast majority of people will not cheat, but there will be exceptions here and there.”

Public officials as nationally prominent as Rhee almost never disavow so strongly statements they made just two days before. But Rhee has been the exception to many rules. She didn’t want the D.C. schools job. When she took it, she ignored many rules of politics, like getting community support before closing schools. That won her enemies, including many voters who decided that she and her patron, Mayor Adrian Fenty, had to go.

I sensed from my talk with Rhee that one reason she misspoke on Monday was that she had not had time to read either the USA Today story or the investigators’ reports, or to probe the weaknesses of test security protocols in Washington and other districts. She would not comment on whether that’s true, or what exactly led her to make the Monday statements, which I think were thoughtless, insulting and, as she said, stupid.

I think everyone should read the USA Today story, although I have a bias. It was part of a four-part series, “Testing the System,” that was the work of 13 reporters and was conceived and edited by my wife, Linda Mathews, who just retired as senior projects editor at that newspaper.

It is good to have Rhee recognizing the problem of test security because none of us in this national debate — not people like me who think testing is important and useful and not those who think we depend on it far too much — know quite what to do. We have to be able to assess student achievement relatively quickly so we can identify which students and which schools need more help, but we also must guard against relying too much on test results that can be distorted.

Henderson has given us a good start by releasing the reports submitted by the test security company, Caveon, on the D.C. erasures and by asking the D.C. inspector to find out what caused so many answers to be so mysteriously changed from wrong to right. The inspector general must do a full investigation, including interviews with the children who were tested and their parents.

Rhee defended the Caveon investigation, saying she commissioned the group “because I was told they had the best reputation for such work and were the best alternative we had at that time.”

But I found the resulting reports shockingly thin, accepting whatever school officials told them with no follow-up and almost no skepticism.

We have to take this seriously now. We must make sure that whatever tests we use, they are as accurate as possible about what our kids know, and especially what they don’t know.