Same data for two 1040s; two different tax liabilities.
I am asking for reactions to the following situation before submitting my 1040 to the IRS. Every year I prepare my 1040 using two different methods to check on consistency of results. I either use two widely used tax software packages, or I use one package and have an organization like AARP prepare the second return. In each previous year the calculated tax liability has been the same for both approaches.
For 2023 I chose to use TurboTax and FreeTaxUSA. The two outcomes were nowhere close to each other. TurboTax calculates that I owe the Feds another $400 on top of the $20,000 that I have already paid them while FreeTaxUSA calculates that I am owed $30. I have carefully checked the numbers that were entered into each template, and they were identical. The aggregate value of each of these categories are the same: ordinary dividends, taxable IRS distributions, taxable Social Security, taxable pensions, and early withdrawal penalties.
However, there are differences in the taxable gains (line 7 of IRS 1040) and taxable interest (line 2b) categories though the sum of these amounts are the same for both templates (ie, taxable gains + taxable income (Turbo) = taxable gains + taxable income (FreeTax)). The difference occurs because FreeTax puts the value of all capital gains (short term, long term, unspecified) into the Capital Gain/Loss line (7) of 1040 while TurboTax adds the value of short term and long term capital gains to the Taxable Interest line (2b). [note: TurboTax claims that only the short term capital gain is included in line 2b but my calculation indicates both short and long term gains appear here]. The short and long term gains come from Form 1099-B and INT-DIV provided by two financial organizations and are entered into the two templates.
I hope that my description is reasonably clear. Is there something that I have overlooked that explains the difference in tax liability calculated by TurboTax and FreeTaxUSA? If I submit FreeTaxUSA and the Feds subsequently find something in error can I be penalized for relying on an established software package?