There are a number of reasons the estimates produced by IPUMS will be slightly different than the information available through FactFinder, but it all stems from the difference in data being used. FactFinder uses the entire ACS dataset whereas the IPUMS data is a 1% representative sample of the population drawn from the ACS dataset. The PUMS files are also subjected to a number of disclosure avoidance measures, which include top-coding certain variables, to prevent individuals from being identifiable in the data. Also, for certain estimates the Census Bureau will employ a variable called ADJUST which you can read more about here. Finally, since you are specifically looking at Households, you may want to exclude group quarters from your analysis using the GQ variable. Group quarters and vacant units are excluded from the HHINCOME universe, but it is important to be aware of how they are being handled in your particular analysis. Because they are not in the universe for HHINCOME, group quarters will be coded 9999999, if this value is not set to missing or if the group quarters are not excluded, this could drastically increase an estimation of household incomes as each group quarters would be counted as having an income of $9,999,999.
I hope this helps.