On the representativeness of exit polls II: the 2018 gubernatorial elections

In my own state, Georgia, Kemp won the governor's race with 50.2% of all votes cast, to Abrams' 48.8%. Using CNN's exit poll to estimate the results yields the following estimates for the gender exit poll:

Kemp = (.52*.46)+(.49*.54) = 50.4%, so 0.2% high.
Abrams = (.46*.46)+(.51*.54) = 48.7%, so 0.1% low.

For age (4):
Kemp = 49.9%, so 0.3% low.
Abrams = 48.8%, so exactly right.

For age (6):
Kemp = 49.7%, so 0.5% low.
Abrams = 48.0%, so 0.8% low.

Using the Excel spreadsheet I just threw together it is easy to calculate estimates for any race if you have the exit poll results (% of voters in each group and voting results by group). If you do it for the FL governor's race (won 49.6%-49.2% by the Republican Ron DeSantis; CNN's exit poll is here), you get this:

Gender: DeSantis = 49.8% (0.2% high), Gillum 48.7% (0.5% low).
Age (4): DeSantis = 49.9% (0.3% high), Gillum 48.8% (0.4% low).
Age (6): DeSantis = 49.0% (0.2% low), Gillum 48.3% (0.9% low).

For this Age (6) poll, the probable reason the estimates for both are a bit low is that 4% of 30-39 year old voters responded "No Answer" when asked who they voted for. Was this because they voted for a third-party/independent candidate, or did they not vote (for governor at least) at all? Probably a combination of both, but some of them probably just didn't answer even though they actually voted for DeSantis or Gillum.

CA: Democrat Newsom won the governor's race in CA with 61.9% of the vote to 38.1% for Republican Cox. Notably, it seems like these were the only two candidates on the ballot because of CA's weird runoff system, so 100% of all votes were for one or the other. Because some exit poll results have at least 1% of respondents refusing to answer we have to make all percentages based on the % of respondents who did answer (100%-the % who didn't answer).
Using CNN's exit poll for that state yields these estimates for the results (all errors are +/- based on value of (estimated-actual)):
Gender: Newsom 60.9% (-1%), Cox 39.1% (+1%)
Age (4): Newsom 60.8% (-1.1%), Cox 39.2% (+1.1%)
Age (6): Newsom 60.7% (-1.2%), Cox 39.3% (+1.2%)