Chuka Umunna is a single person, as much as his baffling political transgressions over the past few years might suggest otherwise. 
I agree with the bulk of your argument, though: A LD surge that takes out BoJo is a particularly deranged fantasy. A net loss for the LDs compared to their current total but not their 2017 result is basically guaranteed, as far as I'm concerned.
Meant Gyimah, embarrassingly enough
I find it hard to imagine them not making it to the mid-20s or so on present numbers; if they win 12 seats on 7%, they should win 24 seats on 14% (which is on the lower end of current polling) assuming their vote stays as efficient as it was in 2017; if anything given that their gains are among strong Remainers, who tend to be a pretty geographically concentrated demographic, I'd expect them to get *more* efficient. (I'd bet on ~30 seats or so, I think).
But I can't imagine the world that some Labour supporters/the editors of the New Statesman seem to be suggesting where they're on 14 seats but are taking Esher & Walton, Wokingham, and Chingford & Wood Green. Surely some Lib Dem seats will come in constituencies without prominent Tories running?
I'm sure part of it is that the LibDem's support may he almost comically disproportionate, largely bottled up in quaint and insulated remain constituencies in London--the sort of people who can't fathom supporting Jezza the supposed British Hugo Chavez but also can't fathom supporting a Tory party taken over by what they see as xenophobic chavs taking away their remain-y multinational corporate European dream. Obviously, there are only a precious few places like this, but it is entirely possible that the LibDems may do quite well in them even as they do poorly nationally.
This is an FPTP election; the more bottled up in specific communities support is, the more seats you get out of it.
More broadly, if they've doubled their support (they've gone from 7% in 2017 to an average of 14.6% in the last 5 polls) they should be at least doubling their seat count. In general in the FPTP system votes become
more efficient the more of them you have, and more efficient the more concentrated they are; the Lib Dem vote is expected to both double and become more concentrated to parts of London. So you should conservatively expect them to win double as many seats as they did last time around (which would be 24), unless you're assuming that their gains are concentrated in places they did
poorly in 2017. Unlikely; it doesn't sound like anyone's ramping Thurrock.
Based on constituency polling (which I know is terrible, but comparing like with like here), their surge in West London might be
very seat-specific, with them within striking distance in certain seats (C of L&W, Kensington, Wimbledon), but very far from winning others, like Chelsea.
Starting with the formation of the Alliance, here's how extrapolating from what percentage it took to win a seat last time around do for the Lib Dems:
1983: Toy model predicts every 1.25% wins a seat --> 20 seats. Alliance wins 23.
1987: Toy model predicts every 1.10% wins a seat --> 21 seats. Alliance wins 22.
1992: Toy model predicts every 1.03% wins a seat --> 17 seats. Lib Dems win 20.
1997: Toy model predicts every 0.89% wins a seat --> 19 seats. Lib Dems win 46 (!).
2001: Toy model predicts every 0.37% wins a seat --> 49 seats. Lib Dems win 52.
2005: Toy model predicts every 0.35% wins a seat --> 63 seats. Lib Dems win 62.
2010: Toy model predicts every 0.35% wins a seat --> 66 seats. Lib Dems win 57.
2015: Toy model predicts every 0.40% wins a seat --> 20 seats. Lib Dems win 8.
2017: Toy model predicts every 0.99% wins a seat --> 7 seats. Lib Dems win 12.
Broadly this basically always works except in 1997, when there was a large amount of pro-Lib Dem tactical voting, and in 2015, when there was a large amount of anti-Lib Dem tactical voting. (Clegg's 2010 campaign does not seem all that great by this metric either; by contrast Farron seems like he exercised a successful refocus on their heartlands at a time when the party
should have lost seats and was facing extinction).
This model shows that Lib Dems won a seat for every 0.62% of the vote they won in 2017, so unless they become less efficient -- and
winning more support almost always makes you more efficient -- you're going to forecast, on present polling numbers, 14.6%/0.62% --> 23.5 seats as a prediction you expect them to slightly outdo.
