Ladybird crystal set question on biasing the transistor

Sorry for a probably basic question, I'm just getting into electronics. I'm learning as I go, but don't really know how to debug many problems.

I'm building the old Ladybird crystal set, as seen here:
http://www.mds975.co.uk/Content/trfradios02.html

I've replaced the OC71 transistor with an NTE 102A.
You build the radio in stages, and with the first transistor installed I can hear many stations and it sounds great. The next step has you go back and bias that first transistor by adding 33k and 150k resistors. As soon as I do this, I lose all station selectivity and it just becomes one big mess of a single station. Any help on how I can go about debugging this?

The webpage describes these as regenerative receivers. To quote, "The regeneration control has to be very carefully set for each station to extract best performance."
I found this to be true when I experimented with regeneration. It was crucial to adjust it to find just the right spot. Then the station suddenly burst through like magic. However the adjustment was so finicky that I got tired of doing it repeatedly.

Do you operate your set in this manner? Did you definitely achieve the regeneration effect? And when you add the amplifying stage do you also adjust the regeneration?

-------------------------------

In regard to selectivity, this is associated with Q of your LC tuner. You want the LC tuner to respond to one broadcast frequency at a time. Q is affected by impedance. Or to put it another way, the amount of current going through the LC tuner. The amplifying stage has something to do with impedance. However we may not be able to predict how this or that transistor will behave.

The webpage names a few transistors as being suitable for the radio. However it doesn't say what makes those types special. Since you use a different transistor, evidently it does 'something' to the signal. Nevertheless it should be possible to adjust values of its neighboring resistors, so you maintain high Q.

It's hard to say whether you should reduce, or increase, the 33k and 150k values.
Did you pick a transistor designed for rf frequencies?
Or perhaps it would help if you select a transistor with greater gain.

There is also a trick you can use, where you can put an NPN stage after a PNP stage (and vice-versa). This trick reduces parts count, by eliminating need for capacitors, and eliminating a bias network.

Back to the old-school days..
I remember and never forget that I realized a SW transmitter with ECC83+EL6 pair in 1975.
Long time ago..

Oh yes...I remember my first 2m receiver/transmitter using a single ECC88...calling CQ I got an answer: ...get a #@%& out of here, this is avionics band