MEPS data observation numbers are not matching

I was working with ipums MEPS dataset and strangely I found that the frequency for each year is not matching the number of observations every year provided by MEPS and documented on IPUMS website: https://meps.ipums.org/meps/userNotes_size.shtml . Please let me know why I am getting this result.

The frequencies on the IPUMS MEPS sample size page are correct, but may be different than expected. In populating the sample size table, IPUMS MEPS followed AHRQ’s convention of showing the valid sample size. That is, those cases that were “key and in scope.” The MEPS is a sample of households that responded to the NHIS. Key persons are those individuals who were associated with an NHIS responding household at the time of the NHIS. People who are “in scope” are those who are members of the civilian, non-institutionalized population.

Finally, note that this is the formal definition, but in cases where the data did not match the definition, IPUMS MEPS opted to show the numbers AHRQ displayed in their table here. IPUMS MEPS is in beta and, as we move the website out of beta, we will likely opt to show the number of cases rather than the “key, inscope” population since it appears to be inconsistently identified by AHRQ in the MEPS data over time.