The assumption doing the most work in your original model was absorption pace, not rent rate. You were $4 off on the ask, which hurts, but the 90-day fill assumption on a storefront that had already been dark eight months is where the model broke. Prior vacancy duration is probably the single most predictive variable for how long the next vacancy runs, and broker comp sheets don't surface it. They show achieved rents on deals that closed, which are the ones that worked.
A few structural points on how this plays out in mixed-use specifically. Ground-floor retail in a two-unit configuration carries concentrated risk in a way a larger retail strip doesn't. One vacant bay is 50% vacancy in that component. When you underwrite the second bay, you're making a binary bet, full or empty, with nothing in between to cushion a miss. The residential above carrying debt service is the only reason this is recoverable rather than critical, and that's worth naming clearly: the structure worked the way mixed-use is supposed to work under stress.
On the TI you mentally counted as sunk but never spent, that's a real accounting distortion. The cash is still in your account. The model was wrong, and the model's error cost you the opportunity cost of vacancy duration, but that budgeted TI is not gone unless you spent it elsewhere. Worth separating those two losses cleanly before you assess total damage.
For the tax prep LOI that went elsewhere: what was the gap? If it was rent, TI structure, or lease term, that's recoverable on the next prospect. If it was parking, frontage, or something physical about the space, that's a different problem and it changes how you should price the vacancy going forward.
What does the current pipeline look like for that bay, and do you have a read on why the tax prep office chose the other location?