You don't mention if you are logging your lifting time in MFP as Strength training - because the Fitbit is actually going to be underestimating that calorie burn since it's not step based.

TDEE is probably even higher yet. Except for one thing that can foul it up.

So you workout 1-3 hrs weekly, but is your 40-45 job time really sedentary? Which is what those TDEE tables assume.

So long as you log non-step based stuff in MFP and let it sync over to Fitbit increasing your TDEE correctly, then trust the Fitbit over your guess of 5 rough levels.

Now - why you may not be able to trust the Fitbit yet.

It bases all non-moving time on your BMR, from gender, age, weight, height.

But a more accurate BMR based on bodyfat could be far different and lower. And indeed if sedentary, could be much smaller TDEE than it's estimating.

Also, if gaining weight on 2100 for 3 weeks, but losing fat, you are eating above maintenance or TDEE.

How much have you gained in 3 weeks?
And what was exact average eating amounts for that time?

You said never a low cal eater, but what level do you think you used to eat prior, and was it the same exercise then?

A minor 10% deficit is also small enough to wipe out with bad food logging, do you weigh all foods and only measure liquids?

The other reason the Fitbit can be off, and that's if you have a suppressed metabolism and maintenance compared to average, it won't know that.

Sure the steps and activity seem about right? Like you don't see a busy 30 min during your drive to work?

First, TDEE, Total Daily Energy Expenditure, in other words maintenance.

If you are gaining weight outside expected reasons for water weight gain, you are eating more than that.

So 3 lbs in 1 weeks would indeed have to be some water weight, so you were eating in a diet prior to that, or there would have been no glycogen stores to even top off to cause it.
It could also be some invalid weigh-in days, low water stores first weigh-in, high the second, giving false water weight gain outside true expected water weight gain.

Was each weigh-in morning after a rest day eating normal sodium levels, not sore from last workout?

Don't worry about Syncing Fitbit in that case, because it does mess up manually entered values. You do want to use it to determine best estimate of TDEE, then manually create goal on MFP based on that.

But, on Fitbit site you should indeed go add in your strength training time and use their calorie count to replace what the Fitbit saw - because it didn't see enough.
That will give better estimate of TDEE, which is why you have it afterall, right.

Then see how the week looks for TDEE.

Good job logging on weekends now, all prior data can be thrown out the window without knowing what was eaten there.
At least it sounds like not too bad of a deficit, just enough.