Using that same method for 1912:
(snip)
Wilson 291, TR 131, Taft 94, Debs 14, Chafin 1
House Seats only: Wilson 210, TR 111, Taft 96, Debs 17, Chafin 1
The differences in our lists
Alabama: Wilson 7, TR 2, Taft 1
Arkansas: Wilson 4, Taft 2, TR 1
Colorado: Wilson 2, TR 1, Taft 1
Georgia: Wilson 9, TR 2, Taft 1
Indiana: Wilson 6, TR 3, Taft 3, Debs 1
Iowa: Wilson 4, TR 4, Taft 3
Louisana: Wilson 6, TR 1, Debs 1
Maine: Wilson 2, TR 1, Taft 1
Maryland: Wilson 3, TR 2, Taft 1
Missouri: Wilson 7, Taft 5, TR 3, Debs 1
Texas: Wilson 13, TR 2, Taft 2, Debs 1
Virginia: Wilson 7, Taft 2, TR 1
Washington: TR 2 Wilson 1 Taft 1 Debs 1
One source of differences is probably due to my method.
Any candidate who gets half of the percentage needed for one vote gets votes counted for the distribution. The votes counted there are used as the total sum. Then the percentages come from dividing the votes by that total.
Then in cases where rounding everybody up creates too many votes then the percentage furthest from being rounded up is rounded down.
Virginia, 1912
Wilson: 6.67
Taft: 1.72
Roosevelt: 1.61
Since there's 10 votes and rounding up creates 11 votes, then only Wilson and Taft get rounded up and Roosevelt is rounded down.