Two important considerations when doing family-level analyses are 1) only select one record per family unit and 2) exclude group quarters. Since family level information is attached to every person record, it is possible to count every person as a family, which means that a family of 4 people would be miscounted as 4 families of 4 people (since the number of family members is stored on each person’s record). Since group quarters (GQ) are treated differently it may be important to exclude them from your analysis. Also, in multi-year ACS samples all income and dollar amount values are inflated to dollars of the last data year (so, if you are using the 2009-2011 ACS, all income figures are in 2011 dollars). This means that you should only use one multiplication factor (that of the last data year in the sample) when adjusting FTOTINC to 1999 dollars. Finally, it may be important to note that POVERTY is based on pre-top coded income values that, this means that there will likely be discrepancies between the variable value and calculated values.
I hope this helps.