The Czech Republic (or Czechia) is facing the second wave of COVID-19 epidemic, with the rate of growth in the number of confirmed cases (among) the highest in Europe. Learning from the spring first wave, when many countries implemented interventions that effectively stopped national economics (i.e., a form of lockdown), political representations are now unwilling to do that again, at least until really necessary. Therefore, it is necessary to look back and assess efficiency of each of the first wave restrictions, so that interventions can now be more finely tuned. We develop an age-structured model of COVID-19 epidemic, distinguish several types of contact, and divide the population into 206 counties. We calibrate the model by sociological and population movement data and use it to analyze the first wave of COVID-19 epidemic in Czechia, through assessing effects of applied restrictions as well as exploring functionality of alternative intervention schemes that were discussed later. To harness various sources of uncertainty in our input data, we apply the Approximate Bayesian Computation framework. We found that (1) personal protective measures as face masks and increased hygiene are more effective than reducing contacts, (2) delaying the lockdown by four days led to twice more confirmed cases, (3) implementing personal protection and effective testing as early as possible is a priority, and (4) tracing and quarantine or just local lockdowns can effectively compensate for any global lockdown if the numbers of confirmed cases not exceedingly high.